[Pascal Boyer] Tradition as Truth and Communicatio

Cambridge Studies in Social Anthropology

General Editor: Jack Goody

68

TRADITION AS TRUTH AND COMMUNICATION

Tradition as Truth andCommunicationA cognitive description of traditional discourse

PASCAL BOYERFellow of King's College, Cambridge

CAMBRIDGEUNIVERSITY PRESS

Published by the Press Syndicate of the University of CambridgeThe Pitt Building, Trumpington Street, Cambridge CB2 1RP40 West 20th Street, New York NY 10011 USA10 Stamford Road, Oakleigh, Melbourne 3166, Australia

© Cambridge University Press 1990

First published 1990Reprinted 1994

British Library cataloguing in publication data

Boyer, PascalTradition as truth and communication: a cognitive description of traditional discourse.1. Society. Role of customsI. Title303.372

Library of Congress cataloguing in publication data

Boyer, PascalTradition as truth and communication: a cognitive description oftraditional discourse / Pascal Boyer.

p. cm. — (Cambridge studies in social anthropology. 68)BibliographyIncludes index.ISBN 0-521-37417-01. Intercultural communication. 2. Cognition and culture.3. Language and culture. I. Title. II. Series.GN345.6.869 1990303.48 2 - dc20 89-7285 CIP

ISBN 0 521 37417 0

Transferred to digital printing 2004

CUP

Contents

Preface page vii

Acknowledgements x

1 Conserved world-views or salient memories? 12 How to think with 'empty' notions 243 Criteria of truth 464 Customised speech (I): truth without intentions 615 Customised speech (II): truth without meaning 796 Customised persons: initiation, competence and position 947 Conclusions and programme 107

Notes 121

Bibliography 131

Index 138

Preface

Although the bulk of anthropological literature is about traditions andtraditional societies, there is no such thing as a theory of tradition in socialanthropology. As Shils puts it (1981: vii), a book about tradition is verymuch in need of a tradition. That anthropologists do not generally recognisethe need to develop such a theory is in itself a puzzling feature of thediscipline. Social or cultural anthropology began as an attempt to describeand understand exotic societies, almost all of which were traditional; themain theoretical constructions were erected to explain typically traditionalways of thinking and behaving. The functional interpretation of myth andritual and the description of marriage prescriptions as structured exchangewere meant to shed light on mainly if not exclusively traditional customs.Trobriand, Zande, Nuer and Navajo, these names used as landmarks inalmost every anthropological discussion or speculation, are all names oftraditional groups. When trying to uncover the whys and wherefores ofstrange customs, Malinowski or Evans-Pritchard certainly got the familiar,if exasperating answer 'we do that because we've always done so', 'becausethat's the way we do it here', 'because our fathers told us to', and so on. Thisdoes not concern the legendary pioneers only; although modern anthro-pology is not ill at ease in modern urban environments, it still is muchmore geared to describing and explaining traditional ways.

Not the problem

In the anthropological literature the characterisation of tradition is generallyconsidered both self-evident (everyone knows what a tradition is) andimmaterial (the way you understand it has no empirical consequences). HereI shall try to show that, on the contrary, it has both difficulties andconsequences. One of my aims is to show that there is a problem, that therepetition or reiteration of tradition implies complex processes of acquisition,memorisation and social interaction which must be described and explained.

Like most anthropological categories, 'tradition' is not a propertheoretical concept. Diverse institutions are lumped together under the label

Preface

'traditional', mainly on the basis of their resemblance to certain prototypicalcases. There is no a priori reason to think that the institutions thus labelledhave any scientifically relevant properties in common, in the same way asthere is no reason to think that the ordinary distinction of trees and shrubsis relevant to biology; scientific categories often cut across common sensetaxonomies. In other words, a theory of traditions like the one outlined heremay not be a theory about all the things usually called 'traditional'. This isjust because ordinary language, including its anthropological variety, issloppy, and we must try not to be. According to Shils (1981: 12), the ordinaryconcept of tradition ' includes all that a society of a given time possesses andwhich already existed when its present possessors came upon it'. Surely, ifthat is tradition, then there cannot be a scientific theory of tradition, just asthere cannot be a theory of rectangular objects or a theory of white animals.There is no reason why all the objects included should share properties otherthan the trivial. Understood in that way, the category is useless (as Shilshimself, I hasten to add, makes clear in his book).

It is therefore important to make clear what this book is not about. It isnot about written traditions, i.e., the conservation, transmission and exegesisof written texts or 'scriptures'. The argument here is based on examples ofpurely oral communication; although the interaction described can be foundalso in literate societies, as I will explain, it is found there in the form of anoral interaction. The question, whether the institutions described here andthe conservation of texts have anything in common, is an empirical one,which surely cannot be solved unless we have a reliable description of whathappens in both domains; but we have no such description so far.

Even granted this restriction, there are still many things called ' traditional'which I will not examine. Indeed, I will focus on a very narrowly defined setof institutions, in which important truths are supposed to be expressed bylicenced speakers, and attention-demanding ritual gestures are performed byspecific actors, all this being accomplished with reference to previousoccurrences of the same statements and the same gestures. My assumption isthat such situations of social interaction share certain important propertiesand therefore constitute a proper scientific object.

'Tradition' and psychology

Without opening a long methodological digression, I must mention the fewcommonsense principles I have followed here. A fundamental one is thatthere is no theory of what happens in cultural interaction without somestrong hypotheses about what is happening in the actors' minds. So weshould try to make these hypotheses explicit; if they are not plausible, weshould change them so that the anthropological theories are at leastcompatible, if not altogether congruent, with what psychologists know aboutmental processes.

viii

Preface

There are two possible ways of making psychological claims inanthropological theory. One is to make use of some precise hypotheses inexperimental psychology, to examine their relevance as concerns the specificanthropological data and to put forward new explicit hypotheses whereneeded. From the days of Bartlett and Rivers to modern cognitiveanthropology, many anthropologists have followed this path. Nothing of thekind, however, has been done about the psychology of tradition, which hasbeen approached in the other possible manner, i.e., by considering humancognition as a ' black box' in anthropological theories. The main idea in theblack box approach is that it is possible to make strong claims about theresult of cognitive processes without bothering to examine how actualpsychological mechanisms could bring about such phenomena. In this view,the technicalities of cognition, the workings of the black box, do not reallymatter; what counts is the alleged result, namely the 'conservation' of certainconcepts and beliefs. When a certain property of traditions is observed, orindeed conjectured, a psychological mechanism is postulated, whoseoperation would give precisely such results.

Against this ad hoc approach, it is in fact possible to make precise claimsabout the processes involved in traditional interaction, and to compare themto some important findings and hypotheses of cognitive psychology. Thisresearch programme obviously implies a great deal of speculation, especiallybecause cognitive anthropology has only recently begun to consider thedomain of ' symbolism' and other such complex phenomena of interaction.However tentative and limited, the findings of cognitive psychology allow usto see why some anthropological theories are just psychologically im-plausible. It is often possible to replace unnecessarily complicated modelswith simpler psychological hypotheses. Throughout the book, I will usepsychological findings in this way, as a safeguard against the multiplicationof ad hoc entities and processes. This way of proceeding is founded on twoassumptions which, I suppose, do not need to be justified in great detail: oneis that unnecessary entities should not be multiplied; the other is that oneshould not place too much confidence in anthropological analytical concepts.Most of them are more like ' phlogiston' than ' oxygen'; they do not refer toreal things in the world, and they designate problems instead of solving them.This is why I have dispensed, as much as possible, with such terms as'symbols', 'cultural models' or 'socialisation'. Important (and unsolved)problems are often concealed under these familiar categories.

Lastly, I must point out that the ethnographic examples given in supportof various hypotheses must be treated in the same way as examples in allanthropological theories, i.e., as suggestive illustrations rather than definitiveproofs. Most of the recurrent pattens the theory is based on are well knownto anthropologists. This is why the analysis of all the implications of theory,at each step of the argument, is given priority over the elaboration ofsuggestive ethnographic descriptions. As a consequence, what is proposed at

ix

Preface

the end of the book is a set of general hypotheses, which further empiricalmaterial should flesh out, modify and possibly refute.

Acknowledgements

The ideas developed here are the distant outcome of long conversations withAnne de Sales and Michael Houseman about possible formal theories of the'cultural world'. Notes and references cannot do justice to the inspirationdrawn from constant interaction with them, and with Carlo Severi whoprovided so many inspiring ideas and criticisms. I would not have developedand corrected the original draft without the substantial help and stimulatingenvironment provided by the University of Cambridge. I am especiallygrateful to Jack Goody for inviting me over and helping me in so many ways.St John's College and King's College gave me the means to do the researchand discuss its results. I also benefited from the detailed comments andcriticisms on a first version of this book and several related papers, especiallythose made by Cornelia Sorabji, Nick Thomas, Tanya Luhrmann, DeclanQuigley, Caroline Humphrey, Andras Zempleni, Esther Goody, GeoffreyLloyd and Ernest Gellner.

Conserved world-views or salientmemories?

The 'traditional' cultural phenomena concerned here can be brieflycharacterised by the following features:

(i) they are instances of social interaction:(ii) they are repeated;

(iii) they are psychologically salient.

Let me comment briefly on these abstract phrases, using a prototypical caseof the sort of institutions I will talk about. After his first trance, a newlyinitiated shaman tells an audience of fellow-shamans and villagers of thecontract he has made with the spirits of the jungle. I will call this event'traditional'. This usage of the term is not really contentious. It is important,however, to be clear about what sort of objects we are dealing with;otherwise some crucial (and typical) mistakes may creep in and make thediscussion hopelessly confused.

First, what I mean by saying that I will deal with instances of socialinteraction is simply that we are concerned here with actual events, with thethings people actually do and ethnographers observe, not the thingsanthropologists think should be posited in order to explain what people do.That is to say, the social phenomenon considered here, and labelled' traditional' is the shaman's singing, more precisely the social event of theshaman's singing to a certain type of audience, about a certain type ofmystical experience, etc. Now an anthropological account of this eventshould include many other things, notably a series of hypotheses about theorganisation of people's ideas, the social organisation of the place, people'semotional involvement in the ritual, and so on. In the anthropological usage,both these underlying things and the social events are called 'traditional'.This confusion, as we will see, is the root of some very problematic claimsabout tradition. In order to avoid conceptual promiscuity, we shouldtherefore try and keep separate labels for the phenomena we observe and theunderlying processes we hypothesise. Here the events will be called'traditional'. In the following pages I will try to examine general properties

Tradition as truth and communication

of' traditional contexts' and ' traditional situations'; such terms will alwaysrefer to specific social events.

Let me now turn to a more difficult point. The events described here as'traditional' are repeated events. It is important to understand that this is acriterion of recognition, not a theoretical claim, in the same way as tellingsomeone that 'giraffes' are 'these tall animals with a very long neck' givesthem a criterion to identify what is being referred to, nothing more. So wewill call the event of the shaman's singing 'traditional' because it is arepeated event, because it is performed with reference to previous occurrencesof the same type of social event. This, in practice, is how anthropologistsrecognise that they are dealing with a traditional institution: at people'sconstant reference to past occurrences, and at the resemblance between theseoccurrences. Not to put too fine a point on it, repetition is an observationalterm.

It is very difficult, however, not to make a fatal mistake here, whichconsists in equating the repetition of occurrences and the conservation of amodel. The event of the shaman's singing seems very similar to whathappened when other shamans were initiated, two years ago, ten years ago,and so on. It seems reasonable to surmise that there is some cultural modelof 'shamanistic initiation singing', which seems to be conserved over theyears. Indeed, it may be the most reasonable hypothesis, but we must remainaware of the fact that this 'conservation' is not an observed property of theevents, but a hypothesis put forward in order to account for their actualrepetition. This distinction, however pedestrian, is crucial because if the'conservation of models' is a hypothesis, then it may be false and it must bediscussed. One just cannot take it for granted; supporting evidence isrequired.1

We will not examine all the aspects of repeated social interaction, and thatis why the third criterion is pertinent. The shaman, to return to our example,probably uses a language whose vocabulary, syntax and phonology are verymuch the same as thirty years before. That is not what we want to explain.On the other hand, anthropologists are interested in explaining why theseinitiation journeys should be sung rather than talked about, why it is deemednecessary to induce a trance, or why the whole process is supposed to curesomeone's disease. To borrow a term from G. Lewis's study of ritual (1980),we are interested in the 'attention-demanding' aspects of these institutions.An utterance of a gesture will not be considered traditional if it does notfocalise people's attention more than ordinary discourse or actions.Psychological salience, therefore, is another criterion of recognition.

Again, these criteria do not constitute a 'definition' of tradition. They aresimply used here in order to pick out a certain class of social phenomena;although they are no doubt too vague, I hope they make it possible torecognise the class of institutions the argument will be about. In fact,

Conserved world-views or salient memories

recognition should not be too difficult, since such institutions (myth,'religious' or 'political' ritual, etc.), are the subject-matter of most classicalanthropological monographs and theories so far. Whether it is legitimate togive the term 'tradition' this narrow extension is a purely terminologicalquestion. As long as the reference is recognised, my 'traditional' institutionscould as well be called 'XYZ'; but that would be just too awkward.

The point of a theory of'tradition' (as identified here, henceforth withoutquotes) is to describe the general processes whereby the salient aspects ofcertain phenomena of social interaction are repeated or reiterated. There isno satisfactory anthropological theory to deal with this problem. The aim ofthis chapter is to show that the anthropological answers are unsatisfactory,not only because the solutions are empirically wrong, but also because theproblem itself has been misconstrued. In order to go further, we musttherefore examine and discuss some common assumptions about traditionalactions and utterances. Unfortunately, the 'common' conception, preciselybecause it is common, is almost always left implicit. It is not explained intheoretical essays or general textbooks, but it is pervasive in ethnographicdescriptions and the generalisations based on them. So it takes a little readingbetween the lines to uncover and discuss it. I must admit right now that mydescription of common anthropological views is intended as a springboardfor further speculation rather than a detailed analysis of the discipline'simplicit premises.

The 'common' conception of tradition

When doing fieldwork in a traditional environment, an anthropologist isbound to give some kind of answer to the question, why and how theinstitutions identified as traditional get repeated. And the type of answer heor she gives to that question reflects in the particular hypotheses put forwardabout those specific institutions. Things are not so simple, however, and thequestion of repetition is almost always coupled with another one, whichconcerns the 'cohesiveness' of the institutions considered, i.e., what keepsthem together; to put it in a less abstract way, what is the link between, e.g.,the specific initiation song for shamans, people's utterances about spirits andthe fact that shamans are said by an informant to cure other people? Thereis no a priori reason why the questions of repetition and cohesiveness shouldbe examined together; as we will see, however, the 'common' anthropologicaltreatment of repetition implies a strong hypothesis about cohesiveness, andvice versa.

The question of cohesiveness is answered in a way which reflects pervasiveanthropological assumptions about culture and society in general. The ideais that utterances, actions and more generally the bits and pieces of behaviouranthropologists observe and record, are in fact held together by some


underlying intellectual objects. These are called 'world-views', 'culturalmodels', 'local theories', 'collective representations', etc. There is aconsiderable ambiguity in anthropological literature about: (i) what part ofthese objects is supposed to be in people's minds and what part is thecombination of what different actors know and think; (ii) how implicit orunconscious they are; and (iii) to what extent they are a reality or a modelof it.2 Here I will not examine these ideas at such an abstract level. I will onlydiscuss their consequences as regards the treatment of traditional institutions.The main point of this conception is that traditionl phenomena are linked to,and explained by a set of underlying ideas or representations.

As for the question of traditionality, the common answer is that somegroups and societies are 'conservative' or 'traditionalistic'; they are soorganised that change is ruled out or deemed dangerous or interpreted as athreat to the social order. This is construed as a characteristic either of thesocial groups, the organisation of which inevitably favours the reproductionof past practice, or of the people, who are naturally conservative orfrightened for some reason at the possibility of change. And it is oftenassumed that this in fact is the normal or natural state of affairs in humansocieties; the problem then is to understand how some societies, like modernWestern ones, seem to foster the idea that change is welcome or inevitable.

These assumptions are generally supposed to be unproblematic. They areseldom discussed, or even presented in too much detail. Anthropologytextbooks for example generally gloss over the exact status of collectiverepresentations, or present a few examples of traditional 'conservatism', togive readers the feel of the thing. Anthropological assumptions are thustaken as a matter of fact, rather than as hypotheses which could be discussedin terms of plausibility and heuristic value. To cite but one example, in arigorous and detailed study of anthropological theories of religion, J.Skorupski has no compunction in writing that a traditional believer sticks tohis creed 'precisely because he is traditionalistic' (1976: 204). That anotherwise punctilious author should not mind the tautology is an indicationof how natural and self-evident the idea of' conservatism' or 'traditionalism'is supposed to be. It is no surprise, then, that these ideas are in most casesleft implicit. This makes our discussion much more difficult, as we always runthe risk of creating the easy target of an anthropological strawman. It maybe of help here to consider one of the very few explicit conceptions oftradition, that put forward by R. Horton in a series of papers on thecomparison of African traditional thought and Western science (1967a andb, 1970, 1982).

According to Horton, traditional thought as a whole should be consideredas an attempt to reach a theoretical understanding of the world, essentiallycomparable to scientific theorising. Theories are built in order to explainevents by integrating them into 'a wider causal context'. In traditional


thought as in Western modernity, according to Horton, we can find two levelsof thinking and discourses: that of'primary theory' (i.e., common sense oreveryday notions about the world) and that of' secondary theory' (traditionin one case, science in the other) which is supposed to make up for the'incompetence' of everyday knowledge in certain areas of human experience(1982: 229). Primary theory is built on common sense forms of reasoningwhich seem to be much the same in all societies; its explanations resort to a' push-pull' type of causality, linking observable events which concernmiddle- sized objects. On the contrary, secondary theory resorts to ' hiddenentities' (like waves and particles in Western physics, deities and mysticalsubstances in traditional thought). Horton then proceeds to the distinctionbetween traditional secondary theories and modern scientific ones; twofeatures can serve as fundamental criteria of demarcation. Traditionaltheories are founded upon a ' traditionalistic' view of knowledge, i.e., theassumption that knowledge handed down from former generations isnecessarily better than new adaptations, because it is ' time-tested' (1982:238). Also, tradition resorts to a 'consensual mode of theorising', 'in whichall members of a community... share a single over-arching framework ofsecondary-theoretical assumptions and carry out intellectual innovationwithin that framework' (1982: 229). Alternative theories generate anxietybecause they break this social consensus. On the contrary, Western scientificthought is anti-traditionalistic (more recent theories are supposed to bebetter) and resorts to a competitive mode of theorising.3

Obviously, we are dealing here with a particularly strong formulation ofthe common assumptions. Horton is not satisfied with such vagueanthropological constructs as 'world-views' and 'conceptions'; for him,traditional phenomena are the expression of theories. It is clear in his firstpapers that the term must be taken in a strong sense; traditional thought, justas modern scientific theorising, produces sets of beliefs that are (i) integrated(ii) consistent and (iii) explanatory. Anthropologists however are constantlydealing with utterances and actions that seem inconsistent or paradoxical,and usually demand more explanation and interpretation than they provide.So they cannot be entirely at ease with Horton's strong claims. Hence a longand heated controversy about the format of the ' theories' anthropologistsare expected to find underlying traditional phenomena.4 Here I will not dwellon the intricacies of that discussion, most participants of which shared whatI called the 'common assumptions' about tradition, namely (i) that traditionsare conserved because people want to transmit them unchanged, and (ii) thatthey are held together by some underlying ideas which constitute a generaldescription of the world. These are the hypotheses I will challenge, and Ithink my arguments apply to all versions of these ideas, from Horton's verystrong formulation to the 'soft', watered down version that can be found inother authors.


In the rest of this chapter I will try to show that such an approach replacesa genuine empirical question, about the processes of repetition, withquestion-begging abstract explanations. To return to the example mentionedat the beginning, we want to account for the repetition of a certain type ofsocial interaction, e.g., between a newly initiated shaman, his colleagues, anill person and an audience; we want to know why all this is attention-demanding ; we want to explain why only certain aspects of the interactionare reproduced, while others are forgotten or left aside. These are difficultempirical questions, and the search for general properties of such interactionsdemands a painstaking process of hypothesis building and testing. If, on theother hand, we just assume that all this happens because people 'stick to theirtraditional theories', the whole question is magically eliminated. The'solution', however, does not really hold; in order to see how much it is onthe wrong track, however, and to grasp exactly what is wrong with thispervasive conception of traditions, it may be of help to present a moredetailed ethnographic example.

Fang literary tradition

The example is that of a traditional literature I have observed and studied,that of the Fang of Gabon, Cameroon and Equatorial Guinea.5 The maingenre is called mvet, after the name of the harp which is used to accompanythe singers. The repertoire is composed of extremely long and complex epicstories, most of which narrate the interminable wars between two clans ofmythic heroes, the immortal giants of the village Engong and the mortallineage of Oku, who try to steal their secret 'life medicine', the secret ofimmortality. Numerous characters are involved in the intricate plots; mostof them are fierce heroes who straddle giant iron elephants and throwrainbows or red-hot iron balls at each other; in case of danger they oftenmake a quick escape beneath the earth or above the sky. They slay each otherby the thousand, and cases of magic resuscitation are not uncommon.

The mvet stories are told only by specialised singers, who have undergonea long and difficult personal initiation under some reputed poet's supervision.Becoming a singer is very much conceived of as the equivalent of becominga witch-doctor. The rituals are quite similar, as well as the ambiguousreputation. Both initiations enable one to master the domain of witchcraft,due to a special relationship with the ancestors; but this capacity may alsobe used for other, anti-social purposes, so that most villagers are rather leeryof these uncanny practitioners. A mvet session is an important social event;people from neighbouring villages gather at night in a 'men's house' and thesession usually lasts until dawn. The epic is interwoven with other literarypieces, notably an obscure account of the singer's own initiation, togetherwith anecdotes, jokes, proverbs, etc. Mvet story-telling appears to be one of


the main elements of Fang tradition. In those sessions, it is generally agreed,important truths about such matters as ancestors or witchcraft arecommunicated; these important ideas cannot be reached in ordinarycontexts. Mvet players are among the initiates, those who know about pastknowledge and hidden entities. Moreover, mvet-related knowledge is all themore valued as it is a marker of identity. Although neighbouring groups haveequivalent and often very similar genres, mvet is considered as exclusivelyFang.

The storytelling sessions have all the usual characteristics of a traditionalphenomenon. They are performed by authoritative specialists who constantlyrefer to knowledge transmitted across generations, during long initiationrituals. The picture is, however, more ambiguous if we try to describe the'world-view' or 'conceptions' expressed. One of the striking features of themvet stories is that most crucial notions in Fang discourse are used here ina puzzling way, often contradictory or paradoxical. A revealing example isthe way the ancestors are described; they are mentioned in both the storiesand the narrative of the singer's initiation. Both descriptions are quitecomplicated; the ancestors' unpredictable behaviour brings about suddencoups de theatre in the narration. As for the lyrical evocation of the singer'sinitiation, it is generally so obscure that even the competent ancestor-cultspecialists get bogged down in the intricacies of the poet's adventures. Tocompound these difficulties, both descriptions are strikingly different from,and sometimes incompatible with, what is received as common wisdomabout ancestors. The same can be said about the other conceptual domainsevoked during the mvet sessions; if epic stories and songs convey a world-view, it certainly gets very muddled in the process. This does not mean thatsome kind of conception is not expressed, but that it is certainly difficult forthe audience to represent what it consists of.

In such a context, the idea that some 'world-view' is conserved acrossgenerations is rather difficult to evaluate. The first good texts were recordedin 1959 (see Zwe Nguema 1972) and a comparison with epics recorded in1981, from the next generation of singers, shows little change insofar as thestyle, characters and main plots are concerned. This, however, concerns the' surface' of the institution; it would be wrong to infer that any important4 meanings' have been conserved, since it is quite difficult to describe them atany stage of the institution.

The problems described here are not specific to Fang literature; mostanthropologists have that kind of difficulty when trying to identify ordescribe the 'conceptions' behind instances of traditional interaction. Africantraditional literature and Amerindian shamanism elude such descriptions inthe same way. These difficulties, however, are not only the result ofincomplete ethnographic descriptions. They indicate a more fundamentalproblem, concerning the lack of 'fit' between the cultural phenomena


observed and the predictions of the theory. In the rest of this chapter I willexpand on this point, and try to re-formulate the problem in a way that ismore consistent with the type of data anthropologists actually gather.

Digression: is long-term conservation relevant?

Before proceeding to the discussion of the common paradigm, however, theFang example will allow me to introduce an important distinction, betweenthe phenomena a theory of traditions is supposed to explain and thoseoutside its scope. In the above description I have left aside what someanthropologists may consider the crucial problem about traditionalsituations: is the alleged 'conservation' of cultural material real! that is, dogroups or people involved in traditional interaction really conserve the sameform of interaction over generations? In the 'common'conception describedabove, the question is either avoided or treated in a rather agnostic way.People stick to established 'conceptions' or 'world-views'; what matters isthat they believe those conceptions to be 'time-tested', not that they areactually conserved over long periods. A striking illustration of this pervasiveview is that, although most anthropologists focus on traditional interaction,they seldom study it across time, measuring its changes over long periods;and those who do are led to emphasise change rather than permanence. So isthe permanence of tradition anything but an illusion ? In the course of tryingto describe traditions as empirical phenomena, I have found, to my ownsurprise, that this in fact is one of the least important and difficult questionsposed by traditions. Without anticipating too much on the rest of theargument, I must indicate briefly why the questions of stability and changewill not be discussed at much length here.

What is described as traditional in ethnography consists of actions orutterances which are performed with the guidelines provided by people'smemories of a previous occurrence. For instance, this year's fertility rites areperformed in a certain way because that is the way the specialists and otherparticipants remember the other years' festivals. In the long run, this processmay well result in a gradual and thorough change; it may also result in nearperfect repetition. Now whether things go one way or the other depends onhistorical and ecological factors which are largely independent of the factthat the ritual is traditional.

Let me return to the example of Fang epic poetry. It is traditional in thesense that the essential features of the communicative event are repeated, andthat people's memories of previous sessions provide the standard againstwhich present performance is evaluated. The complex apprenticeshipfollowed by young men until they become fully fledged epic poets is largelybased on the memorisation of whole stories and should normally result in thereproduction of themes and style from generation to generation. Storytelling


sessions however do not happen in a cultural or historical vacuum; therelevance of the epic motifs depends on what representations the audiencecan associate with them or infer from them. When G. Tessmann, the firstreliable ethnographer in the area, transcribed fragments of epic songs in1913, these could be considered as 'war-songs'. Violent inter-clan and inter-ethnic feuds were rife at the time, and epic poetry was meant to inspirebellicose enthusiasm. Two generations later, however, the colonial and post-colonial national order has put an end to traditional warfare. Epic combatsare more obviously 'mythical' than in Tessmann's time. The evocation ofwitchcraft powers, on the other hand, is certainly more central than it usedto be, as the colonial order is widely interpreted as a period of thrivingsorcery. Now the fact that story-telling has changed in content and relevancedoes not mean that it is not traditional; conversely, the fact that it istraditional does not imply that it conveys the same representations and hasthe same effect on successive generations.6

The phenomenon a theory of tradition must explain is why and how itseems so natural to people, in certain circumstances, to take last year's or lastdecade's version of some myth or ritual as the only relevant way ofperforming these actions. This, as we will see, is an important empiricalquestion, which cannot be solved in the usual, question-begging way or withtautological hypotheses. The theory therefore focuses on the process oftraditional repetition, and must leave aside the question whether the processleads to actual cultural permanence across many generations, because thisresult is only partly dependent upon the process. To take a distant analogy,a good theory of reproduction should explain the processes whereby livingbeings generate other living beings which are very similar to them. Onecannot expect the theory to explain the evolution of species, because suchfactors as mutation and adaptation are outside its scope.7

The reason why these questions are confused is that anthropology oncebelieved it could kill two birds with one stone, as it were, namely explain bothtraditional repetition and the historical evolution of societies within the sametheory. This led to the fabrication of such fictions as 'cool' or ahistoricalsocieties, to the description of current hunting groups as similar to paleolithicancestors, etc. Such ideas relied on two confusions: (i) between a form ofinteraction (tradition) and the kind of societies where this interaction seemspredominant (I will return to this point in the last chapter); (ii) moreimportantly for our present argument, a confusion between describing aprocess and describing its output in all possible circumstances. The latterconfusion is the one we must avoid here. What we are aiming for is adescription and explanation of the processes whereby past occurrences of aninteraction are the reference of present occurrences. This phenomenon isextremely widespread, it constitutes the subject matter of many anthro-pological descriptions. It has different long-term results, depending on


extraneous factors; but we do not want to describe these factors here, onlythe process itself.

Problems generated by the common conception

We started with an empirical question; we wanted to know why and howcertain salient forms of social interaction get repeated. We noticed that theanthropological ideas about tradition, however pertinent, do not address ouroriginal problem directly. Traditions as effectively studied are clusters ofrepeated, salient, etc., events. General theories of tradition, on the otherhand, focus on intellectual constructions ('world-views', 'conceptions','models', 'theories', etc.. Instead of dealing with the repetition of actualinteraction, they focus on the conservation of underlying cultural models.Obviously, this is a different question; while repeated events are observed,conserved ideas are hypothesised. The two problems just cannot be confused.

The reasons why anthropologists substitute conservation for repetitionand models for actions is that they have an implicit causal hypothesis.Cultural models cause actual behaviour, and their conservation causestraditional repetition. People repeat rituals because rituals express ideas andpeople think ideas ought to be conserved. Obviously, there are many nuancesand subtler formulations, but the hypothesis is there. Indeed, it has to bethere, otherwise there would be no reason to describe traditional institutionsin terms of underlying conceptions. The causal hypothesis is not reallydiscussed, because it is viewed as self-evident; as a result it is not evenconsidered a hypothesis at all, rather a fact of the matter. Now, if we takeseriously the fact that it is a hypothesis, we must examine its 'cost' andcompare it to rival explanations. Amazingly, these aspects are almost neverenvisaged in anthropological discussions, so that we have a theory oftradition without any examination of what other explanations would be like,therefore without any argument. Here I will first focus on the 'cost' of thehypothesis. The common anthropological claims are not as trivial and self-evident as they may seem. They are 'expensive' in two ways; first they do notfit the data very neatly, so that one has to put forward additional hypotheses,to bridge the gap, as it were. Second, they imply some strong claims aboutthe way people's minds work when processing traditional actions andutterances. Obviously, the fact that a hypothesis is expensive does not entailthat it is wrong, unless we have rival hypotheses which will do the jobwithout the expense. I will try to show that this is precisely the case.

Let me first focus on the cost of the claims, and on the problem of 'fit'mentioned above. The traditional phenomena observed are series of repeatedactions and utterances. Although it seems reasonable to suppose that thereare some underlying conceptions, describing them and establishing the linkbetween them and the 'surface' phenomena are no easy tasks. In general

10


essays about 'tradition', the claims about cohesiveness and repetition seemsimple, almost self-evident; in practice, they lead to worrying difficulties,especially if we take them for the explanation of traditional repetition. This,I will argue, is because some crucial features of the cultural phenomena athand are neglected in the abstract description. It is important to identifythese characteristics, because they may suggest another approach totraditions. By the same token, this discussion will allow me to refine thecursory description of traditional institutions given at the beginning of thischapter.

An important feature of traditional practice is that, in most cases, theactors do not bother to justify or rationalise it. Typically, the anthropologistis told 'we do this because we've always done so' or 'because it must be so,otherwise it would not be proper' or 'because the ancestors told us to do it',and this type of statement is certainly part of the specific intellectual climateof traditional institutions. This of course does not mean that traditionalpractice is without rhyme or reason, but, more precisely, that traditionalthings seem to provide their own justification. Performing a certain ritual, forinstance, is of course justified in terms of practical goals: solving a conflict,healing a person or placating the ancestors. But the fact that the ritual hasto be performed in a specific way, by specific people, does not seem to requireany explanation; it is amply justified by the ritual itself. The mvet epics forinstance could be sung at noon by uninitiated women, instead of at night byinitiated men. But that would not be mvet, for the Fang; why is that?Because mvet is, precisely, something that is sung by initiated men at night.I do not mean to deny that in some places people are eager to explicate,justify and rationalise their traditional rituals. I am only suggesting that suchexplanations are not a necessary condition of such rituals, since (i) they areabsent in many societies and (ii) when they are present, even anthropologiststake them for a posteriori constructions rather than the raison d'etre of ritualaction.

Another important property that is often neglected or misunderstood iswhat I would call the literalism of repetition. People do not seem to considerthat a description of a certain event, however faithful, can have the samecognitive effects as the original version. This is especially striking whenapplied to utterances; rather than explaining the point of a certain statement,people tend to repeat it literally and to avoid paraphrase and interpretations.When commenting on a statement made by a mvet singer or a witch-doctor,Fang people are careful to preserve its exact wording, even if a morecolloquial or less ambiguous version is available. This is all the more strikingas 'literalism' is absent from everyday conservations. This literal style isprobably among the intuitive criteria that help anthropologists recognise atraditional institution. It is, however, very easy to muddle the issue by sayingthat people are 'conservative', so that they stick to the very form of the

11


statements, and not only to their content. Again, this may be plausible, butit is an interpretation, not a fact. The point I made above, about repetition,applies here in the same way: a property of people's utterances is not thesame as a common hypothesis that seems to account for it.

A related characteristic is that traditional institutions are pervaded with'event-talk' rather than 'theory-talk'. That is to say, people seem to find theactual traditional utterances or actions more attention-demanding than ageneralising commentary. Informants asked to comment on, for example,whether ancestors are jealous of living people are more willing to refer tosingular situations than general principles; they will use a series of concreteexamples and leave their moral implicit. Every anthropologist knows that itis usually much easier to elicit a description of a specific ritual, performed byspecific persons, than an explanation of its underlying principles or generalstructure. This is not only an artefact of the ethnographic situation.Traditional institutions are generally ' transmitted' without explicit tuition;people may have the same ideas on ancestors as the older generation, butthey did not receive these ideas in the form of generalising 'lessons'. Instead,what they received were countless examples of specific situations or specificproblems. This focus on events is even clearer in cases where an initiationritual is involved. Many forms of traditional action may be performed onlyby people who have undergone a specific initiation; such rituals (on whichmore in chapter 6) provide people with all sorts of salient memories, but theydo not give much in terms of explicit principles about, e.g., ancestors,witchcraft and water-spirits.

This description is no doubt too vague and intuitive. But I hope it makesit clear that the 'common' anthropological idea of traditions as theexpression of underlying conceptions is not as simple and evident as it firstsounds. When applied to concrete cases of traditional practice, it generatesdifficult problems. When confronted with more general properties oftraditions, it seems strangely inadequate. Again, this does not necessarilyimply that the common assumptions are false; it just means that they cannotbe taken for granted.

Things get even worse if we focus on the additional or ancillary hypothesesrequired, in order that the common conception can be applied. For instance,the claims are about 'conceptions' and theories, but people seem mainly tofocus on events, on actual situations of traditional interaction, while generalprinciples are difficult to elicit. A way of bridging the gap is to assume thatthe conceptions we are supposed to describe are implicit or unconscious. Thisis not necessarily an absurd hypothesis, but it generates some new problems.For one thing, it makes it difficult to maintain the second claim, namely thattraditions get repeated because people are conservative. For how couldpeople be conservative about things they are not aware of? In the same way,traditional literalism is puzzling if we stick to the idea that traditions express

12


theories about the world. Usually, people engaged in theory-building aremore interested in their theory's capacity to explain this or that event thanin its exact wording. This could be solved with another ancillary hypothesis,to the effect that the distinction between the content and the formulation ofa theory does not always hold; that is, some people may believe that changingthe words will inevitably change the theory itself. Again, this claim does notseem too absurd, but it entails that people are aware that there is anunderlying theory-so it cannot be unconscious after all. Why then is it sodifficult to elicit? I will not try to solve these puzzles here. My aim is just toshow that the common claims about traditions are problematic. Thehypothesis, that traditional interaction can be studied as the expression ofconceptions and world-views, generates all sorts of difficult problems.8

Repetition and conservatism: a reformed common conception

There are two possible ways of dealing with traditional repetition; it can bedescribed either as an intentional process or as an automatic one, i.e., eitheras something that happens because people want it to happen, or as a processthat is quite independent of people's desires. The 'common' anthropologicalconception, obviously, favours the intentional hypothesis; it claims thattraditional interaction is repeated because people or groups are 'con-servative'. This is often presented as a matter of fact, or even as a universalfact of human psychology, without much explanation. The commonconception, however, would have a hard time trying to explain exactly whatpeople are conservative about. The obvious choice, which would make thetheory at least coherent, would be to say that they are conservative aboutunderlying conceptions and world-views. This, as I said above, would beextremely problematic; if people are conservative about conceptions, thenthe conceptions must be explicitly represented; but there is good evidencethat they are not. We have a set of interactions which get repeated. We alsohave a set of important characteristics like literalism and event-talk, whichimply that people's attention is focused on what could be called ' superficialaspects' of the interaction, as opposed to their 'deep', underlying aspects.

The only alternative left is that people are conservative about these'surface' properties they seem so much interested in. We may thereforesuppose that what are transmitted are these aspects themselves, not theunderlying objects one may hypothesise in order to make the actionsreasonable. In other words, interaction is traditional because people repeatcertain sets of actions and utterances, not because they hold the same'theories' as former generations. This hypothesis seems overwhelminglysupported by the evidence on traditional interaction. As I said above, peopleinvolved in such interaction tend to focus on its surface aspects rather thanany 'deep' properties: on the words of the songs and the gestures of the

13


ritual. People just do not transmit the meaning of a myth or a ritual. Theygive other people the details of the stories, and the precise recipe forperforming the ritual in a proper way. They seldom comment on thetheoretical principles expressed by the ritual action, they rather argue aboutits details. They are therefore interested in literal repetition and findmodifications or paraphrases not quite proper. They tell anthropologists thatit would be a very serious mistake, e.g., to say the spell four times instead ofthree, or to use parrots' feathers for the masks instead of cassowaries', andso on. Their 'conservatism', if there is such a thing, is directed at surfaceproperties, not underlying constructs.

So it seems possible to interpret traditional repetition as the consequenceof this 'conservatism'. The hypothesis, then, is that the cause of traditionalrepetition is people's conservatism about the surface properties of interaction.This hypothesis I will call the 'reformed' common conception. It can befound, implicitly, in many ethnographic descriptions. With the earlierversion, described above, it shares the idea that traditional repetition is anintentional process, that it happens because people want it to happen. Thereis a crucial difference, however. If we admit the reformed conception, then wegive up the idea that traditions should be described as the expression ofunderlying conceptions and world-views. Two important points should bementioned here, if only to avoid premature misunderstanding.

(i) this hypothesis does not imply that 'conceptions' and the like do notexist. One can hold the hypothesis, and still believe that conceptions andworld-views are legitimate objects of inquiry, in all available versions ofthis idea: with the intellectual objects being in the anthropologist'smodel or in the actors' minds, with them being unconscious or simplyimplicit, etc. What the hypothesis says is just that they are not the causeof traditional repetition; neither conceptions and world-views, whateverthey are, nor people's attitudes to them (if they have any), suffice toproduce repetition;

(ii) also, in case the terms ' deep' and ' superficial' should become confusing,the hypothesis is not based on the premise that people's representationsare less 'abstract' than anthropological models. It may be tempting toconsider world-views and conceptions as 'abstract' ideas, whilememories of actual interaction are 'concrete'. This certainly applies toanthropological texts, where actual interaction is used as an example ofthe claims about cultural conceptions. But we are talking here aboutpeople's mental representations. And there is no reason to assume thatmemorising situations involves 'concrete' ideas, while thinking that' spirits are mischievous' is more ' abstract', a point to which I will returnin the last chapter;

(iii) obviously, saying that the cause of repetition is conservatism, i.e., apsychological attitude, does not mean that only psychological factors

14


are involved. By dropping a lighted cigarette, one may cause a forestfire; now the fact that a fire occurs of course depends on many otherfactors: trees being highly flammable objects, some oxygen beingavailable, gravity pulling the dropped cigarette to the ground instead oflaunching it on orbit, etc. But these facts are irrelevant; all we want tosay is that, had the cigarette not been dropped, the forest would not havecaught fire. The theory just claims that, were not some instances ofsocial interaction represented in a specific way, their repetition wouldnot occur.

The 'reformed' conception seems to win on all fronts, as it were. First, itis empirically plausible. Contrary to the idea that people stick to established'theories', which flew in the face of the facts, the idea that they stick toestablished practice seems undeniable. Second, it has a simple, not too far-fetched hypothesis about traditional repetition. Third, it can even devisesome hypotheses to deal with the underlying objects that fill anthropologicaldescriptions. World-views, for instance, could be seen as the consequence ofshared practice. People who learn a language with native speakers will endup sharing their syntax and morphology, even if they have never been'taught' the grammar, and are still not aware of its structure. In much thesame way, people who perform rituals together could end up having the sameconceptions of spirits and ancestors, just because everyone derives these ideasfrom the rituals. So all the important questions seem to be solved in asatisfactory way. Are they really?

Unfortunately, the main hypothesis in the reformed conception turns outto be its Achilles' heel. It is assumed that the cause of traditional repetitionis people's conservatism. To be more precise, this means that people'sconservatism about some surface properties of previous occurrences (words,gestures, etc.) is the cause of their presence in subsequent occurrences ofsocial interaction. There are two problems with this simple claim; first, it isempirically false; second, it is based on a very odd view of human cognitiveprocesses.

The hypothesis is false: it says that what people want to remember isreproduced, and, conversely, that what is reproduced is what people wantedto remember. There are many counter-examples to both claims. First, manysalient aspects of social interaction can be repeated without anyone being'conservative' about them. Conversely, there are countless examples ofpeople being 'conservative' about something, without succeeding inpreserving it.' Conservatism' is neither a necessary nor a sufficient conditionfor the preservation or repetition of salient aspects of social interaction.9 Ifwe want to identify the cause of traditional repetition, 'conservatism' is nota good candidate (I will return to this point presently, and try to explain whyit sounds like a good candidate).

There is another problem with the 'reformed' conception. In the

15


intentional conceptions, human memory is considered as a storagemechanism which is more or less under people's control. People want to storecertain things, and simply do it; there does not seem to be any difficult'technical' problem involved. All 'important' items are conserved, all therest is discarded. This idea of human memory is psychologically im-plausible.10 Yet it is pervasive in anthropology, because it seems congruentto what the actors say; in many groups where traditional institutions aresocially important, accurate memorisation is praised, and having 'a goodmemory' is a highly valued skill. This discourse is of course founded on theidea that memory is a storage device controlled by the person. Now the factthat many people have this conception of their cognitive capacities does notimply that it is a true conception; psychology would be much easier, ifeverything people believed about their minds was true. An elementary fact ofpsychology is that most memory processes are out of the user's control;the latter's ability to modify or improve the storage capacity or fidelityonly concerns a very small portion of what they actually remember. Thememorisation processes involved in traditional repetition are likely to displaythe same aspects. Contrary to what an intentional conception must assume,there is no direct relationship between wanting to remember something andactually succeeding in recalling it; people are forever forgetting crucial dataand remembering irrelevant material. In other words, people certainly havegood reasons to remember certain things and forget others; unfortunately,such reasons are not always the cause of their remembering and forgetting.

So the ' reformed' theory cannot be true, at least in this formulation. Thesurface properties of interaction are repeated, and people sound conservativeabout these properties. It seems therefore reasonable to suppose that theseaspects are repeated because people want them to be 'conserved'. This, infact, is where the crucial mistake was made, and I will argue that the'reformed' conception, together with many anthropological theories, hasbeen led astray by its vocabulary and the implicit assumptions it carries.

People involved in some traditional interaction tend to think that all itssurface properties are crucially important. So, given that previous occur-rences of the interaction serve as a constant reference, they are likely to focuson exact repetition and reiteration. Now describing this as 'conservatism' or'traditionalism' is wrong for many reasons. These are ordinary languagewords which carry all sorts of common beliefs, and could therefore preempta precise description of the phenomenon at hand. This is precisely whathappens here. 'Conservative' or 'traditionalistic' people, in ordinarylanguage, want social interaction to stay the same either because they findchange unbearable, or, in a slightly less crude way, because they think anypartial change will contaminate the whole; see, e.g., people who want theChurch to stick to its rites because otherwise everything will become anobject of argument, and belief will disappear. So if one is a ' conservative',one is very likely to display some form of literalism about social interaction.

16


This, however, does not imply that the converse is true, that is, that aliteralist behaviour necessarily implies some form of 'conservatism'. Indeed,there are many forms of literalist behaviour that have nothing to do withconservatism at all. Computer science teachers will typically insist that theirstudents, when writing a programme, use spaces, indentations andpunctuation marks in the precise manner they have been taught. Obviously,there is no conservatism here, just the (true) belief that any change of thiskind will preclude successful computing. So one can be ' literalist' and insiston literal repetition because one thinks the desired effects will be broughtabout by the surface properties of the object. This is precisely the attitude ofpeople involved in traditional interaction. As I mentioned above, traditionalliteralism means that one takes, e.g., the precise gestures of a ritual, theirorder, the fact that only certain persons may participate, the allocation ofroles during the ceremony, as the essential aspects without which one is notperforming the ritual at all. To take another example, it means that theprecise succession of episodes is essential in a story, otherwise it is not thesame story at all. Now if one performs rituals in order to achieve certainresults (which, after all, is the point of most rituals) or if one tells certainstories because they contain important truths, one will be led to insist onaccurate repetition and to worry about the smallest changes. One's behaviourin such circumstances will be much more like that of the computer scientiststhan that of the political 'conservatives'.

To sum up: the reformed conception assumes that 'conservatism' is thecause of traditional literalism; but if there is any causal relation here, it is verylikely to go in the opposite direction. It is because people consider the surfaceproperties of traditional phenomena as essential that they tend to favourtheir exact repetition. Describing this as 'conservatism' is very misleading;even if people display some 'conservatism', this cannot be the cause ofrepetition. Lots of things get repeated without people being conservativeabout them; indeed, lots of aspects of traditional interaction get repeatedbecause people do not remember anything else. In such cases it is memoryitself that makes the selection, a point the implications of which I willexamine presently.

Let me now summarise the state of the argument. The common conceptionof traditions relies on two main assumptions: (i) that traditional interactionis structured by underlying conceptions and world-views; (ii) that itsrepetition is caused by people's conservatism. Assuming that people areconservative about underlying conceptions is just untenable. If, on the otherhand, we say that they are 'conservative' about surface properties, wetherefore abandon the first claim. But then we get another problem: theintentional view of repetition is based on a distortion of ethnographic factsand a strange view of human memory. If a hypothesis generates so manyproblems it should be discarded, unless it is the only one available.Anthropological discussions often convey the impression that conservatism

17


about underlying conceptions or conservatism about literal execution are theonly possible explanations of traditional repetition. But that is not the case;another reasonable hypothesis can be put forward. In the following pages Iwill show that it is less expensive, in terms of additional hypotheses andimplicit psychology; in the rest of the book I will argue that it explains more.

Repetition and the representation of interaction

How can we explain traditional repetition without having the same problemsas the common and reformed conceptions ? Since most problems come fromthe intentional view of repetition, it seems reasonable to begin by examiningthe alternative, that is, the idea that repetition is an automatic process, outof people's control. The implication is that, given certain conditions, someforms of social interaction are such that they will be reproduced, while otherswill just disappear. The problem, obviously, is to describe the 'conditions' inquestion. Let me first observe that the idea of an automatic process, whichconverves certain cultural traits and discards others, is not especially new; infact, most evolutionary models of culture are based on the hypothesis ofsome such process, adaptive or otherwise, which 'selects' certain culturaltraits. The analogy with natural selection is constant, although it is notalways clearly understood (Ingold 1986: 33ff).

Here we are not concerned with these long-term processes, but with themore modest task of explaining why an occurrence of a certain socialinteraction displays surface features which are similar to previous occur-rences, and how the previous occurrences provide the guidelines forsubsequent performances. Obviously, the device that performs the selectionis human memory. The intentional view of traditional repetition implies, aswe saw above, a strong and implausible claim about human memory. Anautomatic view will imply some such claims, too; it will imply, for instance,that memory should be considered as a. filter or some kind of sieve.11 Humanmemory imposes definite constraints on the material processed. It cannotstock just any format of representations. Furthermore, it does not retrieve alltypes of representations with the same success. Some types of ideas are easierto store and retrieve than others because of the built-in specifications ofindividual memory, so that they are more likely to be conserved andtherefore to be traditional.

The hypothesis seems to provide a simple explanation of repetition; theaspects of interaction which get repeated are the memorable ones. Stated inthese terms, however, the hypothesis is true simply because it is tautological.What is repeated in a certain culture must have been memorised, and whatis memorised must be more memorable than what was forgotten. That it ismore memorable is demonstrated by the fact that it is better memorised. Theexplanation is bound to remain circular, unless some psychological

18


hypotheses specify the general constraints brought to bear on the ideasmemorised. To take but one example, it is perfectly possible to specify theconstraints that make certain stories more memorable than others. This typeof inquiry, initiated by Bartlett, is based on an experimental study of thememorisation of stories. By varying the format of the experimental material,and observing the resulting differences in memorisation, it is possible to putforward a formal description of the difference between 'good' (i.e.,memorable) stories and 'bad' ones. Not surprisingly, the stories collected inthe field turn out, in most cases, to conform to the structural constraintsdrawn from the laboratory experiments.12 This just implies that in theprocess of transmission, stories are either modified towards a ' good' format,or simply forgotten. The moral of this example is that, if we have a set ofprecise cognitive hypotheses about the memorisation of cultural phenomena,we can have precise, non-question-begging hypotheses about their repetition.We will then be able to interpret traditional myth as made of particularlymemorable stories, traditional ritual as particularly memorable sequences ofgestures and actions, and so on. So the theory of traditions, if it is based onprecise cognitive claims, will be both anthropologically significant (it willpredict things that actually happen, and those it rules out will not beobserved) and psychologically plausible.

There are some problems, however, to do with the assumptions of thetheory. We have two hypotheses here: (i) that the requirements of humanmemory are the cause of effective memorisation (this is the automaticrepetition stance) and (ii) that this is all we need to account for the repetitionof traditional interaction. In order that we can proceed from (i) to (ii),however, another assumption is necessary, namely (iii) that all the repeatedaspects of the interaction are mentally represented. Otherwise the explanationwill not work. In other words, if some aspects of social interaction are notmentally represented, then ipso facto they cannot be memorised. Now if theycannot be memorised, the workings of human memory surely cannot haveany effect on their repetition. So if we want human memory to be the cause ofthe repetition of interaction, than we must assume that the interaction isentirely represented.

Whether traditional interaction is represented, or more precisely, whetherall the relevant repeated aspects are represented, is an empirical question.Some forms of interaction are entirely represented, some are not. To take butsimple examples, playing a string quartet may be a social interaction in whichall relevant aspects of the interaction are represented by all participants; eachof them knows his or her score, the others' score and (roughly) the resultingmusical effect. On the other hand, the daily traffic jams on the motorwaysnear a large city certainly display a recurrent pattern. Explaining therepetition in this case does not imply assuming that the recurrent patterns arerepresented in anybody's mind. Chances are good that there is a continuum

19


between such extreme cases. So where is traditional interaction? Thisquestion is crucial because, if it is at the same end of the spectrum as chambermusic, then it is entirely represented and a ' filter' view of human memory isall we need to account for traditional repetition. If it is at the other end, thenwhatever refined models we have of human memory will not constitute asufficient explanation of traditional repetition. Here I will argue that it issomewhere in between, which of course does not sound like a very audaciousor promising claim. But it is somewhere in between in an interesting way, thatis, the combination of represented and non-represented elements is notentirely haphazard and its description is crucial to our understanding ofrepetition. As a result, hypotheses about memorisation will play a major rolein my account of traditional phenomena, but the 'filter' function ofindividual memory will not be considered the cause of their repetition, onlyan important background condition.

Events, communications and repetition

Let me now abandon the rigid deductions, and introduce some speculativeclaims that will be substantiated only in the course of this book. Most of thetraditional phenomena we want to explain constitute communicative events,i.e., events whose representation by the different participants differ in a non-trivial, principled manner. Communicative events can be generally character-ised as involving different actors, and inducing some changes in at least someparticipants' representations. This of course is both vague and minimal, butthese trivial characteristics of communication have important consequenceshere. If the events considered have any cognitive effects, it is because therepresentations available to different participants are different.

To make this less abstract, let me return once again to Fang epics. Thereis a clear-cut distinction between certain events labelled 'mvet sessions' andother events. The principles of the distinction, however, are not entirelyaccessible to the actors, because a mvet session implies people's participationin many different ways; as a consequence, people's representations of themvet vary from person to person in a non-trivial way. The composition of theaudience, for instance, is not entirely immaterial. When singing about hisinitiation, the poet makes elaborate allusions to various rituals, of which onlyfew listeners have some knowledge. These competent listeners typicallyengage in some informal and highly metaphorical dialogues with the singer;their contribution is indispensable, as is that of the other listeners. Suchdialogues not only imply a strict distribution of ritual knowledge, they alsodisplay it, and this complex interaction between those who know, those whodon't and those who know who knows what, is essential for a good mvetsession (see Boyer 1984, 1988). This is an organised partition; if the listeners'representations were randomly distributed, or if everyone had much the same

20


ideas about ancestors and witchcraft, the communicative effect would not beachieved.

This of course is a limiting case, in the sense that knowledge and itsdistribution are what a mvet session is all about. But the remark would applyto other forms of interaction, and I think (this is where speculation begins)that it would in fact apply to most forms of traditional interaction. Thesalient instances anthropologists study nearly always comprise some specificallocation of roles. To return to the example mentioned at the beginning ofthis chapter, the newly initiated shaman is telling colleagues and othervillagers about his journeys to the world of the spirits. This is a form ofinteraction in which a specific allocation of roles is certainly crucial; the newshaman does certain things qua new shamans, the senior ones qua seniorshamans, and so on. So far, nothing but obvious features of interaction. Thehypothesis I will try to defend, however, goes further. The idea is that: (i)there are differences between the various participants' representations of theinteraction; (ii) these differences are principled', and (iii) they are a causalfactor in the repetition of the interaction. I do not think the first point needsmuch justification; the idea is, for example, that the new shaman certainlydoes not have the same representation of the singing session as the seniorshamans. The main idea, however, is that such differences, beyond individualand other trivial ones, are organised and that their distribution hasconsequences for repetition. That is, the way a new shaman's representationof the singing differs from that of the senior shamans is a crucial element ofthe interaction, without which there would be no repetition.

It is worthwhile insisting a little on this question, if only to distinguishnecessary platitudes and important hypotheses. The platitude is that thereare some differences between the participants' representations. A semi-platitude, as it were, is the idea that these differences are organised, they arenot just the result of personal capacities and interests. The main hypothesisis that these organised differences are among the causes of traditionalrepetition; admitting this has obvious consequences in the empirical study oftraditional interaction. If we think that these differences are unprincipled ortrivial, then we will be led to study, e.g., the shaman's initiation ceremony interms of shared representations about the interaction. We will insist on those'ideas' which are present in most participants' minds, judging the differencesas either unprincipled (there is no organised distribution) or irrelevant (theirpresence does not change traditional repetition). If, on the other hand, wethink that the differences in representations are principled and causallyrelevant for repetition, we will be led to focus on the way certainrepresentations are not shared by the participants. This, I will claim, is theright direction; neglecting this possibility has led anthropology to putforward insufficient accounts of traditional repetition.

To make things more intuitive, there is all the difference in the world21


between explaining a string quartet's performance and explaining a game ofcards. In the former case, all you need in order to explain why these peopleproduce organised sound, not random noise, is a description of sharedrepresentations about the score and each player's part in it. If you want toexplain the point of a card game, you need a set of hypotheses about bothshared information (the knowledge of the rules for instance) and the preciseway some of the information is not shared (the knowledge of one's hand).Anthropologists have generally neglected the possibility that some forms ofsocial interaction may be closer to the latter type, except in cases where theprincipled differences jumped at them, as it were, for example in cases wheresecrecy and mystery were the actors' main preoccupations. Such cases standout because people are aware of the differences and their organisingprinciples; my claim is just that, in many other forms, possibly all forms oftraditional interaction, there are organised differences of which the actors arenot necessarily aware, and that they play a major role in the process ofrepetition.

Conclusion

Here I have tried to discuss at some length the 'common' claims made abouttraditional repetition. At least one point should now be clear: although theyseem self-evident, almost trivial, these claims are immensely problematic.They fail to constitute a coherent, empirically significant account ofrepetition. The confusion lies, not in the solutions put forward, but in theterms of the question. The actual repetition of occurrences of socialinteraction is replaced with the hypothesised conservation of world-views.The fact that traditional material is necessarily memorised is just ignored,and human memory is considered as a neutral, user-controlled storagemechanism. In lieu of empirical hypotheses, what is thus offered is a series ofquestion-begging devices.

It is possible, however, to go beyond these claims, and to treat traditionalrepetition as the object of precise, explanatory hypotheses. So far I have onlyargued for the possibility of such a theory, without making any substantialhypotheses. I have thus replaced the imaginary 'solutions' to a largelyartificial problem, with a real, empirical problem which has no solution sofar. Certain forms of social interaction lead to the repetition of their surfaceaspects. The crucial claim is that, all else being equal, the cause of therepetition is to be found in people's representation of these surface aspects;more precisely, in the way these representations are distributed amongparticipants.

The only way to substantiate this claim is to put forward a series ofspecific, empirically significant hypotheses about various traditional inter-actions, and then to measure to what extent these aspects or their

22


combination contribute to the repetition of the interaction considered. Thisis what I will do in the following chapters. I will consider tradition as a typeof interaction which results in the repetition of certain communicative events.That tradition is a form of interaction, that it concerns communication andresults in repetition are all rather obvious features, which are usuallyobscured by vague 'theoretical' statements about conservatism and othersuch properties. In this book I will try and account for several generalcharacteristics of traditional interaction. I hope it will at least show that, asI said at the beginning, the study of tradition is a matter for explicit,substantial hypotheses rather than vague conceptual deductions.

23

How to think with 'empty5 notions

The translation or gloss of specific cultural categories is an important aspectof anthropological description. It is therefore appropriate to begin oursurvey of traditional interaction and its psychological conditions by someremarks on concepts and categories. Describing local categories is moreoften than not a difficult task, for reasons both trivial and important. Someterms designate institutions which are just absent from the anthropologist'sworld (e.g. 'potlatch') so that the gloss must comprise some description ofthe referent; such difficulties are easily overcome. In this chapter, however,I will focus on the thorny problems posed by some fundamental categoriesof traditional ritual. They seem crucial in the description of whole systems ofrepresentations and interaction; yet their interpretation in anthropologicaltheories is extremely problematic.

The analysis of local categories is usually considered an excellent approach,at least as a starting point, in the description of social interaction. Obviously,the description of social interaction cannot but mention the categoriesactually used by the participants; there is simply no way to describe, e.g., theinteraction of shamans and their clients without considering their notion of'spirits' or 'souls', or some fertility rituals without a notion of 'ancestors',as the case may be. Not to put too fine a point on it, categories should bedescribed because their use is part of the interaction described. People usethem, argue about them, and so on. There is a second, more interestingreason why anthropologists examine local categories, namely, that theirrepresentation in people's minds includes some important assumptions aboutthe entities designated. To put it crudely, concepts seem to carry implicittheories, or some elements of the theories. In order to describe Fang ideasabout death, it may be useful to start with a gloss of the term bekong,(roughly) translatable as 'ghosts'. By providing a detailed, fine-grained glossof the term, we will inevitably evoke some important local ideas about deathand life and bodies and minds. This surely will not be enough, but it is a goodstart.1

It may be important to examine the implicit assumptions in this study oflocal categories, which sometimes leads to implausible theoretical statements.

24

How to think with 'empty' notions

Remember that anthropologists, in order to describe traditional interaction,must take some hypotheses about (i) what representations people have, and(ii) the way these representations make it possible for them to participate inthe interaction. The description of categories is therefore crucial, because (i)they are necessarily the object of some mental representations, and (ii) suchrepresentations cannot be just individual fantasies. It is generally assumedthat local categories are mentally defined, and that their definitions are sharedbetween users. There does not seem to be anything controversial (or indeed,anything worth dwelling on) in such assumptions. In normal conditions, onemay assume that a speaker uses terms like 'giraffe' and 'hairdresser' becausehe or she has some representations of what giraffes and hairdressers are; onemay also assume that people can use 'giraffe' and 'hairdresser' inconversations because they share a good part of these representations. Inother words, if we are looking for shared mental representations, categoriesare the domain in which we can safely assume we will find them.

In this chapter I will focus on cases where the analysis of local categories,based on these assumptions, meets difficult problems. In such cases, twoalternatives are open: (i) we may judge that the description of the sharedrepresentations is wrong or incomplete, should be further refined, etc.; (ii) orelse we must assume that there is something wrong with the very idea ofshared mental definitions. I will present an ethnographic example, someaspects of which are in fact very common in traditional contexts. I will thenexamine to what extent such cases lead to a revision of the commonassumptions about categories and their use.

A Fang category and a general problem

The Fang notion of evur is generally glossed in the anthropological literatureas 'witchcraft substance' or 'witchcraft organ'. Evur is a central notion inFang ritual and in the beliefs concerning witchcraft, medicine, and moregenerally, individual capacities.2 Only some people have evur; this enablesthem to get rich, to avoid misfortune or 'send' it to others, to kill people bywitchcraft, to become epic-singers and witch-doctors, etc.; on the otherhand, people fall ill or die either because 'their' evur is attacked by someoneelse's, or because they are without evur and therefore defenceless.

The domain of evur extends beyond 'witchcraft'; evur is supposed to beinstrumental in all varieties of personal achievement and in every event thatis not related to the system of lineage hierarchy. There is a strong oppositionin Fang discourse between the 'village' order, organised by kinship ties, andthe domain of the 'forest' where evur is dominant. This opposition is alsothat of day and night, order and disorder, cooperation and selfishness, etc.Such personal achievements as becoming a famous epic-singer or a powerfulorator are always interpreted as the result of some 'nightly', evur-related

25


action. The assumption is that successful people are beyem, i.e., evur-bearers;their evur sucks up other people's blood, ruins their plantations, andgenerally undermines their endeavours. Success is always acquired throughsomeone else's demise, and it is interpreted as the irruption of evur powersin the otherwise fair and peaceful village world.

The notion is intrinsically ambiguous. Describing it as an 'organ' isinsufficient, though sometimes pertinent. The Fang used to perform post-mortems, when they thought a person had died as a result of his or herinvolvement in witchcraft activities. When a polypus was found in theintestines, it was identified as the evur of the deceased; but the absence ofsuch a salient trace was not a clear proof of innocence. The term does notdesignate the 'organ' itself, which is only a manifestation of evur; nor doesit denote the special capacities. Specifically, it denotes whatever gives somepeople those special capacities. Analysing this notion is difficult becausepeople have extremely vague views on what evur exactly consists of and themechanisms of its action. People are quite definite, on the other hand, onsuch topics as the type of effects evur has, the social relations that revolvearound evur-related actions, the dangers of dealing with evur-bearers, and soon; but they cannot say what evur is.

Notions like evur are extremely common in traditional interaction. Almostevery anthropologist has, in the course of fieldwork, come across such a term,which is extremely ambiguous, yet is placed at the very heart of traditionaldiscourse. In anthropological descriptions, such terms are often translated as'vital force' or 'mystical energy'. Some people are said to 'have' it, whileothers do not. Informants will readily claim that the ' x ' in question is a good,evil, necessary, powerful, dangerous thing, they will easily describe the socialrelationships and ritual acts connected with the 'x ' , the persons supposed toknow about such things, but they seem unable to define what the 'x ' is. Suchnotions constitute a limiting case for anthropological hypotheses. Althoughthey are of constant use in the traditions concerned, their exact meaningseems to remain obscure or inscrutable to the speakers themselves.

This combination of extreme salience and apparent vacuity constitutes achallenge to the common anthropological ideas about traditional categories.We started with the seemingly plausible idea that categories imply shareddefinitions. Here we are dealing with notions which do not seem to be definedat all, yet are constantly used in a traditional interaction. Indeed, one couldcontend that there is a direct relation between the vagueness or even vacuityof the terms and their use in traditional symbolism. Even if we do not takethis as a causal relation, the correlation is striking. In most traditionalcontexts there is some notion of that kind; and it is usually among thefundamental notions, without which there can be no description of theinteraction concerned.3 But then the idea of shared definitions as the basis oftraditions is clearly threatened. So we must examine in more detail howanthropology tries to cope with such odd categories.

26


Two anthropological approaches

There are, ideally, two ways of dealing with local categories which seemimpossible to define, while maintaining that categories are the object ofshared definitions. A first strategy consists in saying that the categories areindeed undefined, but that they are an exceptional case. Alternatively, onecan assume that the notions in question are not undefined, but that theirdescription is incomplete, because anthropologists have just been looking forthe wrong kind of definition.

The first approach was defended by C. Levi-Strauss in his analysis of theMelanesian notion of mana. From the 1910s to the 1940s, this notion was alocus classicus of anthropological discussions. The notion was used as aparadigm of magical thought; the very notion of mana is the point ofdeparture in Marcel Mauss's famous Esquisse dune theorie de la magie. Thispurely theoretical discussion subsisted for several decades on its abstractlevel, without much consideration for the new data gathered. In a generalsurvey of the theories about mana, Keesing (1984) showed that anthropo-logists consistently ignored the linguistic or discursive evidence, and basedmost of their interpretation on a substantive use of the term {'the mana"),which simply does not exist in the languages concerned.4

Discussions about mana had somewhat subsided when C. Levi-Strausswrote his 'Introduction a l'oeuvre de Marcel Mauss' (Mauss 1951, and inLevi-Strauss 1987), in which he argues that notions like mana are universal('une forme de pensee universelle et permanente') and offers a simplelinguistic explanation. Mana-concepts are taken as central cultural symbolsnot in spite of but because of their semantic vacuity. They must be consideredas 'a zero symbolic value', to be compared with phonological 0 —features(1987: 64). They can also be compared to algebraic symbols, which are inthemselves devoid of meaning, but can be precisely used to denote anynumber. Levi-Strauss offers yet a third analogy; mana-concepts resemble theFrench words true and machin (or the English 'stuff'), generally used todenote any object for which the speaker does not find an accurate name ordescription (1987: 55). For Levi-Strauss, this common usage gives us the keyto the symbolic elaboration on mana notions. The argument can be summedup as follows:

(i) the construal of the world implicit in linguistic categorisation does notentirely fit with human knowledge of the world;

(ii) hence the need for a special signifier which encompasses whateverknowledge ' stick out' of the linguistic framework;

(iii) this 'empty signifier' denotes virtually anything, precisely because it ismeaningless.5

This ingenious explanation is consonant with a general structuralist idea ofcategories as forming a 'grid' of 'signifiers'. Terms like true and machin, and

27


traditional categories like mana, are all-purpose signifiers that designatewhatever 'sticks out' of the structural grid. I will not comment on the grandclaims about human thought and language; their vagueness would make thediscussion difficult in any case. But we must focus on the specific hypothesismade about traditional categories like 'mana' or 'evur', namely that they are' empty notions' which denote just whatever one has no word for.

If we try to apply this idea to any particular case, it becomes clear that thetrue analogy just flies in the face of the facts. Traditional categories areprecisely not used in that way. They are not used as the joker in cards. First,the languages in which they occur generally have some other means to copewith 'signifieds without signifiers', for things one has no word for. Mostnatural languages do provide some 'joker cards', but these are neverconfused with the traditional categories we are dealing with. To take but oneexample, a Fang mechanic who has no Fang word for ' differential gear' callsit dzom ete (this stuff), and certainly not 'evur'. Conversely, it is obvious thattraditional categories are used in a very precise way. Intuitive rules govern theuse of such terms, and the ethnographer's job is to make them explicit. Ifsuch rules did not exist, if the categories actually denoted anything, then itwould be in principle impossible to tell a serious use of the term from meregibberish. But the ethnographers' informants are usually able to make sucha distinction, which suggests that there must be some limits to the use of thecategories.

Although ingenious, Levi-Strauss's true analogy is clearly wrong; in orderto go beyond such explanations, however, it is important to examine theassumptions on which the explanation is based. The main idea is that thestudy of local categories should be directed at language, conceived as anabstract system of interrelated symbols, not at the way language isrepresented and used by speakers. Seen at that abstract level, there may wellbe some resemblance between mana and true. All the empirical objections tothe true analogy concern the representation and use of the categories. Theintuitive rules the ethnographers try to uncover concern the use of the term.Hence no solution to this question can be found unless one considers, notonly the position of the terms in the vocabulary, but also their use in people'sutterances.6

Let me now turn to the second alternative, in which it is assumed thatterms like 'evur' are problematic because their description is incomplete.Anthropologists report that it is very difficult to elicit any definition of suchterms; the apparent 'vacuity' seems an essential aspect of the notions.Without challenging this, it may be possible to argue that anthropologists aremistaken about the status of such terms. A typical example of such anargument is the ' neo-intellectualist' conception mentioned in chapter 1. Inthis framework, tradition as a whole is considered as an attempt to reach atheoretical understanding of the world. Horton thus compares the gods andspirits of African cosmologies to the 'hidden' entities of modern scientific

28


explanations, such as waves and electrons. Both kinds of concepts are usedto place events in a 'wider causal context', and to achieve 'explanation,prediction and control' of certain contingencies. The comparison seems toapply even better in the case of mana notions, although Horton does notdiscuss them; mana or evur or other such ' things' could then be consideredas theoretical principles, and that would explain why they are impersonal andcausative entities. No one can really explain what evur is, no more thanmodern physicists would explain what electro-magnetism is. But to supposethat such a hidden 'thing' exists and brings about certain effects is the onlyway to account for many events. The hypothesis is part of a theory, and inHorton's framework traditional categories are names which designateprinciples described or explained in certain theories about the world.

There are obvious advantages to this interpretation, especially in contrastwith the structuralist claims. The intellectualist paradigm makes it possibleto take into account the obvious fact that there are some constraints on theuse of these categories. The explanation is that the terms cannot be used as'jokers' because the implicit theory fixes their reference; not everyphenomenon fits into the theory. A set of theoretical principles determinewhat is and what is not an electron; in much the same way, for a'neo-intellectualist', there must be a traditional theory which explains what isand what is not evur or mana. There are some obvious problems, however,with this type of 'explanation'. First of all, a limited problem of conceptrepresentation is ' solved' by creating a much bigger problem elsewhere, inthe representation of theories. Not to put too fine a point on it, thearguments developed in chapter 1 should suggest that it is not absolutely self-evident that people entertain 'theories' about such things as evur. Theutterances they produce about this topic are mostly about singular situations;they do not display or even imply consistent explanatory principles. Theoriesabout evur, if there are such things in the minds of Fang people, areunconscious and not really consistent. So we have to admit that, if these ideasare called 'theories', then we should waive the requirements of consistencyand explicitness usually applied to theories. Having 'a theory of evur' willthen mean having organised thoughts about evur.7

This, however, is a self-defeating argument. In this discussion, the onlyadvantage of the idea of traditional ' theories' was that it seemed to explainhow people can have a precise use of a term without a definition. Now if thetheory is vague or inconsistent, how could it put any constraints on the useof the term? If people's 'theories' about evur, for instance, are inconsistentthen they are of no help in using the term; they do not provide any guidelines.The idea of traditional 'theories', in the explanation of traditional notions,leads to the same dilemmas as in the explanation of traditional repetition.The claim is either strong (the term ' theory' is taken in its clear, usual sense)and implausible, or watered down to fit the data and consequently trivial.

29


There is a common point in the structuralist and 'intellectualist'approaches. Instead of considering actual uses of a notion, these approachesfocus on abstract.objects, either the system of' signifieds' or the traditional'theory'. As a result, it is implied that the use of terms like 'evur' is simple,and can be derived from the abstract objects considered. In the structuralistapproach, making an utterance about 'evur' only involves some knowledgeabout the language. In the intellectualist model, it only involves knowing therelevant theory. In order to go beyond these problematic hypotheses, it maybe necessary to examine the way such terms as 'evur' are used in actualcommunication. Against the models presented here, I would claim that wemust distinguish between several interrelated 'registers' or 'styles' ofdiscourse. These distinctions, sometimes explicit, are never ignored by peoplewho use the terms. My account will be based upon the Fang data, but I thinkmany ethnographers will recognise situations they often deal with.

Evur and discourse registers

There are many types of discourse about evur, which is a very common topicof conversation among the Fang. The first and most important source ofethnographic material is people's informal conversations and assertions,what I will call common discourse. At the beginning of this chapter Imentioned some properties of evur; all these were taken from commondiscourse. To take but one example of the contexts of common discourse,people often evoke, in casual conversations, the alleged increase of witchcraftmurders which is seen as a consequence of the colonisation and thedestruction of former ancestor cults. On such occasions, people seldom failto mention the general characteristics of evur. It has 'neither back norstomach', i.e., it is ambiguous, no-one can really grasp it or know the truthabout it. Evur is a 'thing of the forest', it belongs to the uncanny world ofthe bush and the night, away from the order and daylight of the village. Some' selfish' people use evur, and it is a pity one cannot tell them from harmlessvillagers. When an evur-bearer is asleep, he or she in fact goes to mgbel, aforest village where all witches meet and fight. These fights typically result inillness, as a person's evur is weakened or wounded in mgbel.

These assertions are very general and abstract. Few singular examples aregiven, and the origin of this knowledge is always attributed to the ancestors:'the elders knew all about evur', 'they told us about that', etc. There are noformalised circumstances for this kind of speech, neither is the audience'restricted' in any sense. Anyone can hear what the 'common wisdom' hasto say about evur. However, no one can draw any definite conclusions fromthis type of discourse, which is often vague and contradictory. The conceptionof evur that emerges from these casual statements is that anything is possibleas far as evur is concerned, so that any definite straightforward statement

30


should be indefinitely qualified. The speakers themselves always assume thatthey are not knowledgeable enough to be precise and definite. Manyassertions are thus followed by the speaker's self-evaluation: 'this is whatI've been told, but after all I'm not a ngengang (witch-doctor)', 'I tell youwhat I've heard, but who can be sure about such things?', and so on. Socommon discourse can be characterised as (i) generalising; (ii) not related toany precisely defined source of knowledge; and (iii) willingly unconclusive.

These features are virtually inverted in a second type of discourse aboutevur, in gossip about witchcraft. This second form only occurs in the contextof private conversations. The adressee is not, of course, a 'general audience',he or she is chosen carefully. Gossip never uses general assertions; on thecontrary it is focused on singular cases and the only generalisations areimplicit. What is even more important, it is always assumed that the speakerhas a definite interest in transmitting a certain version of the events. This iswhy people constantly remind each other not to trust people who engage ingossip, although people actually do it as often as possible, and importantdecisions are based on information gathered in such conversations.Witchcraft cases are sometimes brought before the weekly customary courts;when pleading their cases, many people must resort to information acquiredthrough gossip. They often try to present it as direct evidence, the origin ofwhich they do not want to disclose; for instance, they 'know' a certainperson has performed a certain ritual, but they will not clearly state that theysaw it. Every time gossip information is mentioned as such, it is instantlydismissed as 'empty talk' or 'lies'; it is rejected on formal grounds. On thecontrary, any assertion that is presented as personal testimony is eitheraccepted or argued against on account of its content.

Gossip is just as inconclusive as common discourse; in every single case ofwitchcraft one can hear many incompatible versions through gossip, and onecan very well relate each version to the speaker's own interests. Furthermore,it is virtually impossible to use knowledge acquired through gossip in anycontext other than gossip itself, because this would result in personalaccusations. In short, gossip is (i) very definite, (ii) centered on singular casesand (iii) of no use in the contexts in which truths are supposedly expressed.

Most common statements about evur, either private or public, general orparticular, are supposedly unreliable. There is, however, another register inwhich definite statements can be found, namely that of the 'experts'. Theseinclude not only witch-doctors but other kinds of healers, and diviners aswell. These people have to make statements about evur, whenever they arerequested to give a supposedly competent opinion on a certain case or event.The statements can be made in private or in public. It is common, forinstance, for people to organise a session of divination shortly after a death,and this session is public.

On such an occasion an expert often resorts to general principles which can31


be found in common discourse, but he also makes a definite statement aboutthe singular case. For instance, a diviner says: 'you all know that evur is ofthe forest, is an evil thing. Those who engage in things of the forest eventuallyhave to pay dearly for it,' etc., However, these pedestrian remarks are but aprelude to the 'real' statement, about the problem in hand. In the same way,a witch-doctor usually proceeds from abstract commonplaces to a graphicexplanation of what is wrong with a certain person, on which side of thefamily the origin of the problem is, etc. In such utterances, there is no clearconceptual connection between the general principles and the singular case.Although the expert presents his conclusions as a mere deduction from thepropositions of common discourse, it is clear to all laymen that only expertscould come to these conclusions and make such statements; this impressionis reinforced by the amount of details mentioned by the expert, as a simpleexample will show.

In a Fang village where I worked I once heard that a young boy I knewwas seriously ill, and I was requested to give some ' tablets', which proved asineffective as those provided by the dispensary. When all other possibilitieswere exhausted, the family took the boy to a famous specialist of evurwitchcraft, who asserted that the illness had begun during a recitation of epicstories I had organised. According to the healer, it should have been noticedthat there were many ants crossing the house where the story-telling sessionwas taking place, and also that the epic singer had put out his tongue severaltimes, for no obvious reason. The final diagnosis was that the boy's evur wasstrong and malevolent, so that he could not stand the presence of anotherevur-bearer, the epic singer in the case at hand. This also explained thelatter's 'agressive' reaction. Everyone found the explanation quite sat-isfactory ; ants and anthills are commonly associated with witchcraft and, forthe Fang, sticking out one's tongue is an agressive and uncanny gesture.

The experts know how to apply the general principles (e.g., about evur,ant-hills, etc.) in a pertinent way, a capacity which is guaranteed by theirspecial initiation. The problem with this interplay of general principles andspecific cases is that the experts' utterances cannot be generalised, except bythe experts themselves. The statements are memorised literally, with all thecontextual details of the case. For all these reservations, one must admit thatsuch diagnoses are an important source of reliable information about evur.To put it briefly, the experts' utterances are (i) definite, (ii) focused onsingular cases and (iii) fairly reliable.

Discourse and shared representations

Let me now try to draw some general conclusions from this ethnographicexample. Although some notion of'witchcraft power', more or less similarto 'evur', can be found in many traditional contexts, I will not pursue this

32


point here. The hypotheses I want to put forward concern the discursiveproperties of the term, the way its use is distributed in different discourseregisters. In the Fang case, we have three main types of discourse about evur:common discourse is generalising but inconclusive, gossip is definite butunreliable, expert discourse is definite and reliable. This rough division,based on the type of information that is conveyed and its reliability, iscongruent with a division between types of context. Obviously, many aspectsof this tripartite division are particular to Fang culture. There is a formalproperty, however, which seems to be a general feature of traditionalinteraction, and therefore deserves more attention.

Although I have drawn a division between three types of discourse, thereis an obvious difference between gossip and common discourse on the onehand, and expert discourse on the other. The difference concerns thereliability of the information conveyed. While gossip and common discourseprovide listeners with relevant material, expert utterances are supposed toconvey true information. As I mentioned, this information is only about asingular situation, it is nonetheless conceived as guaranteed truth. That thereis a difference between unreliable registers on the one hand, and the truth-delivering one on the other, is clear in the difference of cognitive salience.Gossip and common discourse utterances, although they are often extremelyrelevant and interesting, are never the object of so much inference andcomment as expert discourse. Also, expert utterances are not easily forgotten.They are quoted and commented on long after the corresponding situationhas changed. They focus people's attention much more than other forms ofdiscourse.

This division, I will claim, is an important general property of traditionalinteraction. The fundamental categories of such interactions are used inmany types of discourse. There is, however, a principled difference betweenregisters in which reliable information is conveyed, and the rest. In theanalysis of traditional notions, one can make use of two types of data: theinferences that can be drawn from ordinary conversations, allusions, etc., onthe one hand, and some salient utterances, made in specific contexts byspecific speakers on the other. Whatever the idiom in which the distinctionis expressed, and however subtle the differences made within each domain,there is a sharp division between registers of discourse in which speakers aresupposed to refer to what the entities really are, and registers for which thereis no such guarantee.

This general hypothesis, taken on its own, is not really controversial. Mostanthropological descriptions mention this type of distinction, betweencontexts in which truths are produced about the fundamental categories, andcontexts in which information is conveyed without any guarantee of truth.Although uncontroversial, the idea has some important consequences. As Isaid at the beginning, the main reason why categories are supposed to give

33


access to local 'conceptions' is that one assumes there must be shareddefinitions of the categories. The idea of a division between registers,however, makes it necessary to modify this seemingly trivial hypothesis.

Reliable, guaranteed utterances are not made by everyone; the experts, intraditional contexts, are precisely identified speakers. Now this implies thatsome aspects of the representations concerned are, precisely, not shared by allmembers of the group. More importantly, it implies that these aspects arecrucial to any description of the interaction. To return to the Fang example,an analysis of the shared aspects of the representation of 'evur' would notgive any insight into the social interaction concerned. The representationsshared by everyone are the basis of common discourse, which, as I said, is notespecially salient. If, on the other hand, we want to understand the use of'evur' in salient interaction, then we must focus on expert utterances; butthen we are dealing with a type of context in which the interlocutors havestrikingly different representations about the notion, as we will see presently.The difference between expert discourse and other registers cannot beexplained as a difference in definitions, so that we must revise the secondfundamental hypothesis in the anthropological treatment of traditionalcategories.

The acquisition of a 'mystical' notion

The three types of discourse about evur are obviously interrelated. Gossipassertions are based upon some general premises that can be found incommon discourse, and they often mention cases that have been investigatedby an expert; common discourse is often a result of what people have learnedthrough either gossip or experts' assertions. But these different types ofdiscourse are not simply juxtaposed, they are ordered, both logically andchronologically. Expert discourse is conceived as something that goesbeyond common talk or gossip. For the Fang, to be an expert is to have'other eyes to see the hidden resemblance of things', i.e., something morethan other people. Moreover, the levels of discourse are ordered chrono-logically, as one can become an expert only by going through a longinitiation; hence people who become witch-doctors or diviners have handledgossip and common discourse before acquiring their expertise. As aconsequence, the process of acquisition is crucial for our description ofpeople's representations about the notion. It is an initiation process, thedetails of which vary according to the specialised activity one wants to reach.But the general framework is essentially the same.

Let us take the example of witch-doctors or ngengang. The initiation is astrictly individual process. It begins when the ngengang-to-be is taken ill, orreceives other signs which mean that he or she has an evur. But evur in itselfis ambiguous and consequently dangerous even for the bearer. It must be

34


' shaped' (akoma) so as to bring about its bearer's well-being without his orher engaging in witchcraft. This is achieved through a ritual which is the firststep in the initiation process. After the ' revelation' of the presence of evur,the process is variable; one can become a healer's apprentice, and learn manyrecipes and treatments by 'buying' them from one's master. But the crucialsteps in the process are rituals which involve the neophyte's symbolic deathand his encounter with ancestors or forest spirits. In these rituals the initiatehas to explore the world of the ancestors and bring back the byang, thesecrets that will enable him/her to practice as a ngengang.

In his remarkably detailed account of Fang witchcraft and magic, L.Mallart-Guimera gives several examples of such apprenticeship; let mesummarise one of these (see Mallart-Guimera 1981: 158-88). LaurentBikoe's initiation begins with a serious disease, a dysenteric infection whichlasts for several years and brings him close to death. After consulting manyexperts, including Pygmy doctors, he meets Ngono, a famous healer who tellshim he has been 'sent' the 'oil-of-the-ghost', i.e., he is a victim of that specificform of witchcraft which makes one's blood turn black and thick. One night,in a series of dreams, he meets his ancestors who give him marrow seeds toeat, and order him to 'go and cure people', to 'make the byang' (themedicines). Laurent is soon cured of the dysentery, and learns some secretsfrom Ngono, who becomes his master. One day, Ngono takes Laurent to theforest and plunges him in a river. Suddenly, Laurent sees a woman ghost, theone who 'works' for Ngono. At that point Laurent's account becomessomewhat obscure; he seems to make some deal with this woman, to theeffect that she begins to 'work' for him. Every time he is presented a patient,it is the ghost who 'finds' the medicine in the forest.

The healer does not say much about the origin of his technical knowledge,his pharmacopoeia. He admits that, after his dream-initiation, he was not yeta witch-doctor, because he had no special knowledge of the illnesses andplants. He claims to have acquired this knowledge during several years ofapprenticeship with various healers, and also during the period of his diseaseand its (unsuccessful) treatments. The very idea of a medical knowledge thatis not acquired through initiation is alien to Fang thought, and the notion ofa 'byang' encompasses both the recipes and the individual ability to followthem successfully. Initiation to such knowledge is strictly individual; itshould be no surprise that many details vary from one healer's account toanother's. Some essential features, however, are always included in theseaccounts: (i) the ancestor, or the witches trigger the process by sending somedisease or misfortune; (ii) an expert is required to reveal the exact origin ofthe event; (iii) the healer-to-be is cured; (iv) he or she is given the ability tocure people and also to send them illness or death; (v) he or she invariablydeclines the latter offer and is content with 'working for the good', which ofcourse is much less profitable. This process concerns the development of the

35


person's evur, the outcome of which is not necessarily the status of witch-doctor. This is why all experts of evur-related matters, e.g., diviners, epicsingers undergo this type of initiation. During my field study of epic singers,I noticed that their initiation narratives were strikingly similar to those givenby Mallart-Guimera's informants (Boyer 1988: 99-114). They always focusedon a journey to the ghosts' villages, during which they had been given the'byang mvet', the secrets of good story-telling, for which there had been avery dear-and of course unmentionable-price to pay (in Fang culture mostoutstanding capacities are supposedly 'purchased' by giving the ghosts thelife of a parent).

In short, one becomes an expert of evur by going through a series ofpersonal experiences, notably a series of direct presentations of the world inwhich evur is visible. Being able to make definite statements about evur andhaving experienced such direct presentations of the ghosts' world are twoqualities that seem necessarily connected. Indeed, the Fang conceive expertdiscourse as a consequence of such experiences. It is because someone hasbeen in such situations that he or she makes certain statements about evur.The initiation experiences themselves are a consequence of the presence ofevur. One does not decide to go through a certain initiation in order tobecome a witch-doctor, there is no such thing as a 'career' in healing. Rather,one adapts oneself to a situation in which one's evur and its relationship withother people's evur are crucial.

In the acquisition of expert discourse, there are two parallel processes,concerning technical knowledge on the one hand, and the ritual contacts withthe domain of evur on the other. The technical knowledge in itself is notabout evur at all, it is only about plants and their effects, and thecategorisation of symptoms. On its own, this knowledge is supposed to beinsufficient. What makes certain people able to identify the effects of evur isa series of experiences. The expert does not acquire another, more refined4 definition' or ' characterisation' of evur; he or she acquires a repertoire ofsalient memories, which concern singular situations, not abstract principles.Most of these experiences are just unconnected with the propositionsexpressed in common discourse, so that becoming an expert does not involveenriching one's characterisation of the notion. On the other hand, thesememories are constantly used by experts in the identification of specificsituations. In particular, the resemblance between a certain situation at hand,and memories of encounters with the domain of evur, is a guideline in theexpert's work.

This could be said, mutatis mutandis, of most 'mystical' notions used intraditional contexts. Acquiring expert discourse implies undergoing a seriesof specific experiences, often described as an initiation, which cannot beconstrued as the acquisition of a new 'definition' for the notions concerned.Such experiences do not transmit technical knowledge, although the latter

36


may be acquired in parallel. The experiences, not the knowledge, make thedifference between expert discourse and the rest. The case of evur is not reallyexceptional in that sense. Without anticipating the problems I will deal within the following chapters, let me observe that it is extremely difficult todescribe such acquisition processes as the transmission of information. Whatis acquired by the subjects is a series of salient memories, which are then usedas guidelines, together with knowledge, in the evaluation of subsequentsituations.8

Categories and their representation

Here I have tried to show that the idea of shared mental definitions oftraditional categories is not really satisfactory. If we want to focus on therelevant aspects of traditional interaction, then we are led to insist on twoimportant properties of the categories involved, namely, that the differencesin their representation are a crucial aspect of the interaction, and that a goodpart of the representations involved do not consist in mental definitions. ThisI have tried to demonstrate on the basis of strictly anthropological argumentsand data. It is also possible to consider this problem from a more generalpoint of view, and to show that the common hypotheses rely on a strangeidea of the way concepts and categories are represented in general.Traditional categories are, after all, but a variety of categories, so thatanthropological hypotheses about the former should be compatible withwhat psychologists know of the latter. Principles that apply to concepts ingeneral necessarily apply to traditional concepts. Indeed, the idea ofdefinitions which are shared is a commonsense belief about language ingeneral, not just traditional categories.

Commonsense it may be, but it turns out to be generally false. Bothhypotheses, that categories are represented with a definition and thatdefinitions are shared, are simply refuted in many domains of the vocabulary.Before going into the details, let me observe that the vocabulary of a naturallanguage is not a uniform landscape, and that anthropologists should beespecially aware of this. In the models I discussed above, the vocabulary wassupposed to be homogeneous. In the structuralist conceptions, all there is ina vocabulary is a set of' signifiers' with their conceptual counterparts; thereare some cases of lonesome signifiers hovering around, as it were, but theyare exceptional. In the intellectualist models, there are two main varieties ofterms, theoretical on the one hand and non-theoretical (one guesses they mustbe observational) on the other. These models are just too simplistic. Even ifwe limit ourselves to substantives and adjectives, there are many differentvarieties of terms, and the idea of shared represented definitions applies onlyto some of them.

To have a definition, if this term has any meaning, is to have a set of37


necessary and sufficient conditions for an object to be a member of the class.This is possible in the case of'triangle' and (maybe) 'hairdresser', but manyterms must be used without such a list of conditions; furthermore, even incases where it would be ideally possible to list the conditions, the hypothesisthat they are mentally represented by the speakers is just untenable.9 To takebut a single example, natural kind terms cannot be identified by a list ofnecessary and sufficient conditions. Such terms denote living species andnatural substances, e.g., 'giraffe', 'granite', 'oak'. The representation ofthese categories cannot provide a mental definition; no-one can specify thenecessary and sufficient conditions for being a giraffe or an oak.10 Thisexample is relevant in the study of traditional categories for two reasons.First, it shows that mental definitions are out of reach in many commonterms, so that seemingly ' empty' concepts like 'evur' or 'mana' are not reallyexceptional in that respect. Also, the comparison makes it possible to explainsome important properties of traditional categories, in contrast with naturalkind terms.

Terms like 'giraffe' are not represented with a definition. But people haverepresentations that help them use the term in a meaningful way. To take acommon philosophical view, people must have a certain stereotype ofgiraffes, which helps them in identifying giraffes and telling them from lionsor tigers. This may consist of a mental image of a typical giraffe, or of a listof qualities which are usually found in actual giraffes. Here I will not enterthe fierce psychological debates about the way stereotypes are actuallyrepresented, as this is not relevant to our problem.11 The only importantpoint is that a stereotype gives properties which are used to identify singularobjects as members of the class, although they are neither necessary norsufficient. Giraffes for instance may be identified because of their long neck,but exceptionally short-necked giraffes are still, unambiguously, giraffes; theproperty, although it is frequent, is only typical.

An interesting aspect of such categories is that one cannot acquire them bybeing taught the definition of the term, since there is no definition. Terms fornatural kinds are usually acquired by ostension, by being shown someexemplars of the kind. Again, I will not enter the controversies about the wayostensive presentations are processed and contribute to the construction ofa stereotype. All models of cognitive development share the trivialassumption, following which acquiring a natural kind term consists inmemorising some ostensive presentations of exemplars of the kind, and thenextrapolating a mental stereotype of the kind on the basis of thatexperience.12 In other words, subjects necessarily start with particular,singular presentations and gradually build a stereotype which is (roughly)common to all speakers of the language.

The point of these remarks is, by contrast, to highlight a very importantproperty of traditional categories, more precisely of the way expert discourse

38


is acquired. To return to the Fang example, people do not build a stereotypeout of certain presentations of evur. As I said above, such presentations areconfined to some ritual contexts which can be attended to, or fullycomprehended, by experts or 'experts to be'. Besides, people would notreally have to ' build' the stereotype, as it mostly consists of certain vaguestatements that can be heard in everyday common discourse. As a result, thefull acquisition of a term such as evur, i.e., the process whereby one goesbeyond common discourse and achieves expertise, consists of proceedingfrom a stereotype (given by common discourse) to a series of ostensions(provided by the initiation contexts). In the case of natural kind terms,people have a mental representation which gradually converges towards the(roughly) common stereotype. The full acquisition of evur, on the otherhand, starts from the stereotype and gradually diverges towards a series ofmemories of personal experiences.

To recapitulate: anthropology takes the study of local categories as anessential step in the description of traditional interaction; this much isbeyond doubt. The underlying assumptions, however, are not reallysatisfactory. The categories used in traditional interaction are not the objectof shared definitions, and understanding why and how this is the case entailssome strong hypotheses about the interaction itself. Here I have put forwardfour main hypotheses, which I think amount to a general description of theway categories are used in traditions:

(i) there are some principled differences between discourse registers, whichinclude a distinction between reliable 'expert' utterances and the rest;

(ii) the use of the term in different registers is based on different types ofmental representations about the objects designated;

(iii) specifically, the acquisition of expert discourse relies on the mem-orisation of ostensive presentations;

(iv) these memories are not necessarily consistent with the stereotype.Whether these hypotheses are true is an empirical question, which

obviously cannot be answered on the basis of the ethnographic illustrationsused here. But their advantage is to be both consistent with whatanthropologists generally report about the use of traditional categories, andpsychologically more plausible than the idea of shared mental definitions,traditional 'theories' or 'zero-signifiers'. All such claims rely on a descriptionof concept representation and use which is simply untrue of concepts ingeneral.

Special categories or special process?

So far I have used the term' traditional category' in a remarkably vague way,to denote all the notions which seem to be crucial in traditional interaction.

39


The argument, however, relied on the example of strange categories like evur,which are fundamental in a specific tradition, yet are not easily definable andtherefore challenge the common views on definitions. Let me now try to bemore specific, and explain exactly what the hypotheses listed above aresupposed to apply to. We have categories which are the object of a commonstereotype, and of special ostensive designations, irreducible to thestereotype. The problem is, whether we are dealing here with special types ofconcepts, say 'undefinable mystical concepts' or 'mana-terms', or with aspecial process which is applied to any concept in a traditional interaction.

In the former alternative, 'mystical terms' would be considered a specialvariety in the vocabulary, besides such varieties as 'natural kinds', 'sensoryproperties', 'theoretical concepts', and so on. Traditions would becharacterised by the systematic use of such 'mystical terms'.13 If we choosethe other alternative, then we will assume that any type of term can become' mystical' in that sense, provided that there is the right type of interaction.This in fact is a general question, which is recurrent in the analysis oftraditions. One has the choice, between taking the specific properties oftraditions as properties of special objects, or as the application of a specialprocess to objects that can be found in other types of interaction. In eachcase, I will try to show that the second option is far more plausible andsatisfactory.

To return to the question of categories, it is easy to show that thehypotheses presented above apply to many notions used in traditionalinteraction, besides the mystical undefinable mana-like category 'evur'. Thiscan be shown by a very brief examination of another Fang category, that ofbekong.u This word is the plural form of kong, which usually designates aperson's shadow, and is also one of the constituents of the person. Afterdeath the kong leaves the body and wanders around the village, as a rathermalevolent isolated 'spirit'. Funerals contribute to 'anchor' the kong, whichis then supposed to dwell in a village of bekong. The organisation andactivities of such villages are similar to those of the living, although people'sideas about this are very vague. At this stage the bekong are consideredrather benevolent; they are often called betara or bemvam ('fathers','grandfathers'). Lineage and clan cults are directed to them as protectors ofthe village 'order'.

People's typical image of the bekong is that of ambiguous creatures, bothbenevolent and uncanny. This is partly due to the fact that the ideas of' bekong as wandering spirits' and ' bekong as fathers' are both necessarilyconnected (a kong is a malevolent spirit before becoming an ancestor) andnot really compatible. Therefore the bekong are often described asunpredictable or whimsical; it is generally considered better to have as littleto do with them as possible. The bekong can bring about various types ofproblems, from bad crops to serious illnesses. Before becoming 'fathers', the

40

How to think with "empty" notions

bekong are envious of the fun and excitement of village life. Even 'anchored'ancestors are likely to get angry when the living break the rules of normalsocial life, such as clan exogamy. The resulting problems make it necessaryto resort to ritual experts. In much the same way as there are evur experts,some people are supposed to be particularly competent about therelationships between the village and the bekong. These experts usually arelineage elders, and they have undergone the full range of ancestor cultinitiations. Their diagnoses, always focused on singular cases, are the onlyreliable source of information about what ancestors really are. Theseutterances are memorised and cited literally, and constitute a repertoire ofcase-stories which people use to make conjectures about present events.

At first sight, the notion of bekong seems very different from evur. For onething, Fang speakers can always roughly define wljiat bekong are; we are notdealing here with an apparently empty or undefinable notion. The processesinvolved in concept acquisition and representation, however, are very muchthe same. The common stereotype of ancestors does not provide, on its own,any knowledge, and therefore any way of dealing with the entities concerned.The representation of the term includes some definitional features, but theseare clearly conceived as insufficient and the acquisition of the conceptinvolves focusing on singular utterances, which ostensively designate thebekong and their action. People's general representations about the entity areinsufficient if not supplemented, and virtually replaced, by memories ofsingular situations.

Categories and memory

The point of this chapter was to try and describe the specific features of therepresentation and use of categories associated with traditional contexts. Myassumption was that there are such specific features, that categories are notused in traditions in the same way as in other contexts, for example not inthe same way as in usual conversation in the societies concerned. This muchis generally agreed in anthropology, but hypotheses differ a great deal aboutwhat the differences consist of. The structuralist model mentioned above hasa hypothesis about such terms as 'evur' or 'mana', which is problematic fortwo reasons, because it is empirically untenable, and also because it impliesa sharp division between ' empty signifiers' on the one hand (terms like' evur')and ' normal' terms on the other, a class that should include terms like theFang 'bekong'. Obviously, this is absurd; the type of interaction, the divisionof discourse registers and the kind of representations involved seem verymuch the same in the cases of 'evur' and 'bekong'.

In the intellectualist model on the other hand, 'evur' and 'bekong' wouldbe classified together as 'theoretical' terms. This, however, is just asuntenable as the structuralist interpretation, given that people's represent-

41


ations about' evur' o r ' bekong' just do not constitute ' theoretical' principles,unless one bends one's categories so much that 'theory' just means'thought'. Furthermore, the intellectualist idea makes it impossible tounderstand the obvious difference between 'evur' and 'bekong' on the onehand, and terms like 'wavelength' and 'genotype' on the other. But thedifference is obvious; no physicist has to undergo an initiation to say thetruth about 'wavelengths', and no biologist would perform rituals in which'genotype' is ostensively designated. These models provide no hypothesisabout the specific features of traditional categories.

On the basis of the Fang example I have tried to show that the specificfeatures lie in the type of representations entertained, and the way they aredistributed between people who have access to different discourse registers.Traditional categories are not special types of categories; they are justcategories represented and used in a specific way. What is interesting aboutsuch apparently empty terms as 'evur' is that they are a limiting-case, inwhich these features of the traditional process are especially visible. Not toput too fine a point on it, the main hypothesis here is that in the process ofacquisition of a traditional category, the definitional features are replacedwith memories of specific situations, contrary to what happens in theacquisition of most ordinary terms. This process is especially clear in the caseof terms like 'evur', 'wakan', 'orenda' and other types of mystical 'forces',for which there are virtually no definitional features to begin with. But itapplies just as well to ordinarily defined terms like 'ancestors'.

If this is true, then it becomes clear why anthropology is wrong in treatingtraditional 'event-talk' as the expression of underlying theories or con-ceptions. As I said in chapter 1, a notable feature of traditional discourse isthe emphasis on specific situations instead of theoretical inferences, on salientexamples rather than general principles. Theories of tradition have a difficultjob explaining why people do not just make their 'conceptions' explicit. Iftraditional expert discourse is acquired in the way described here, then it isclear that the task is in fact an impossible one. Traditional interaction impliesthat some people supplement a vague and unconstraining common stereotypewith memories of singular ostensive designations. Indeed, their discourse isgenerally centred on such designations and on memories of such contexts.Traditional discourse is therefore based on representations which areessentially different from common stereotypes. As a result, trying to interpretit as the expression of such stereotypes is not only wrong, it is like trying toput round pegs in square holes. Anthropological theories seem to identify as'traditional' something (the expression of common stereotypes), the absenceof which constitutes traditional discourse.

This strange way of dealing with traditional discourse is in fact rooted instrong, and I think misguided, claims about the psychological processesinvolved in the acquisition of culture in general. Here I must briefly introduce

42

How to think with ' empty•' notions

an elementary psychological distinction, between the memorisation of eventsas such and more general representations about the world. Having a set ofrepresentations about what took place at a certain time and location isdifferent from having representations about the way the world is.Remembering the way so and so was dressed yesterday is not the same asremembering the French for 'giraffe'. This distinction, between what wewould usually call 'memories' and 'knowledge', is expressed in psychologyas a difference between episodic and semantic memory. Episodic memory issupposed to store events as such, i.e., with some indication of their location,while semantic memory contains representations which are not linked to aparticular context.15

There is a marked tendency in anthropology to ignore the fact that mostrepresentations about traditions are in fact episodic, occasion-bound. This ina sense may be what makes ethnography such a difficult process. There is asystematic discrepancy between what anthropologists are seeking, namelysome semantic memory data, and what conversations with informantsprovide in abundance: memories of singular situations. The link betweenthese two types of representations is not usually described or analysed inanthropological descriptions; when it is mentioned at all, it is only construedas a one-way process, from semantic memory to the interpretation ofsingular occasions. It is assumed that people use their general representationsabout the world to organise and understand particular situations. Stated insuch vague terms, the hypothesis is beyond doubt. Its extension andgeneralisation, however, gives rise to what Bloch (1985: 21) calls 'theanthropological theory of cognition', the assumption that the individualprocess of building representations from singular situations is essentiallytrivial; it is supposed to consist in the absorption of a ready-made conceptualscheme.

This idea is not only problematical, but plainly inapplicable to traditionalcontexts. Given the preference for 'event-talk' I have mentioned, and theabsence of formal tuition, people must build semantic memory data from thelimited material presented in a series of singular situations; there is simplynothing else to build on. Therefore the processes through which episodicmemories are used to generate knowledge are crucially important toanthropology. Fang people for instance are very rarely told about witchcraftin general terms. In order that they build some representation of whatwitchcraft is, or the type of actions and effects it involves, they must makeuse of whatever singular contexts in which the topic is mentioned. People'srepresentations are thus the result of a generalising mechanism, unless wesuppose that they are communicated through some magical process.

The ' anthropological theory of cognition' suggests that people have aconceptual scheme, which is used deductively to interpret singular situations;they have certain general ideas about, e.g., witchcraft, and can thus interpret

43


a certain situation as a result of witchcraft. This certainly happens, althoughit seems strange that anthropology has not studied the opposite process, i.e.,people using their memories of singular occasions inductively, to modify thesemantic memory and build a representation of the world. The idea thatpeople only make deductions from a 'culture-given' conceptual scheme begsan essential anthropological problem-how people build certain represent-ations of the world from fragments of experience-which in fact must bestudied empirically.

My description of evur and other such cultural categories implies that thefeatures of complex cultural constructs, such as a ' system of beliefs' arepartly determined by the concrete situations and processes of conceptacquisition. The way one learns what evur 'really' means imposes someconstraints on the range of possible beliefs about evur; this simple cognitiveprocess is involved in all processes of concept acquisition. When anthropo-logists suppose that beliefs about mana or evur put some constraints on thesemantic representation of the notion, they probably put the cart before thehorse. People experience situations in which a term is used long before theyhave any beliefs concerning the corresponding concept. This process is all themore striking with seemingly 'empty' notions; here it is clear that, to acertain extent, the circumstances of acquisition constrain people's beliefsbecause nothing else is available to limit the extension of these terms.Obviously, there is no theory or abstract principle, which would determinewhich objects and events are designated; hence people must resort to whatis given, namely utterances of different types, and make inferences about thenotion from what is conveyed by these utterances and what is implied by thefact that they belong to different registers of discourse.

Conclusions

The hypotheses presented here are of course incomplete, as a descriptionand explanation of traditional categories; this is, in part, because it is difficultto account for the use and representation of categories without explainingmany other aspects of traditional interaction, which will be treated in thefollowing chapters. One point in particular deserves special attention. Theinteraction described here implies that there is a principled division betweenthose discourse registers, on the one hand, which simply use a term in acorrect way and convey the common stereotypes about it, and on the other,the registers in which there is supposed to be a description of what the entitydesignated really is. The distinction may be more simply described in termsof guarantees of truth. Certain types of utterances are supposed to conveysome truth about the entities designated; other utterances are not. In theprocess of building some representation of an entity, the inferences derived

44


from the latter type of utterances are certainly weighed in a specific way. Theuse of categories in tradition is crucially dependent on this logical division ofdiscourse registers. A study of traditional concepts should therefore describethe processes whereby specific utterances are interpreted as conveying atruth.

45

Criteria of truth

Questions of truth are both crucial to the study of traditions and usuallyoverlooked in anthropological theories. Anthropologists generally avoidusing such terms as 'true' and 'false' when describing the status oftraditional statements. They use some vague terms, such as 'symbolic' or'symbolico- expressive', 'significant', 'relevant', 'meaningful', etc. There is asharp contrast between this usage and the views of the people concerned. Noone in a traditional society would ask a diviner to make a 'significant' or'relevant' statement about any situation, nor would people be satisfied if ashaman claimed to make 'meaningful' findings in his journeys to otherworlds. What is sought in such contexts is a true description of a situation;indeed, the production of truths is the raison detre of many types oftraditional interaction. Divination comes to mind immediately, as a ritual,the explicit purpose of which is to make a true diagnosis; but it is by nomeans the only such context. All anthropologists have been told byinformants that certain persons knew 'the truth' about certain mattersbecause they had been initiated, or that certain myths contained some ' truth'about the ancestors, while other persons or stories were described asmistaken or misleading. Truth is very much a traditional concern, and theanthropological use of woolly terms ('symbolic', 'meaningful', etc.) is rathermisleading.

The reason why traditional statements are seldom described as true for thepeople concerned, in spite of the informants' emphasis on their veracity, isthat it seems difficult to apply the terms ' true' and ' false' to such statementswithout generating a host of difficult problems. The type of things which aresaid to be 'true' in traditional contexts, and the way people argue for theirveracity, do not seem to square with ordinary ideas on truth and falsity. Hereare but a few examples of the problems that may arise:

(i) in a traditional context, people sometimes admit the truth of a sentence,without admitting the truth of its paraphrase. They judge a certainutterance 'true', but are doubtful about some direct deduction from thecontent of the utterance. This is an aspect of the ' literalism' problem

46

Criteria of truth

mentioned in chapter 1: the words of, e.g., a certain story are said toexpress a 'truth', but another, logically equivalent formulation is notperceived as conveying a truth;

(ii) in the evaluation of an utterance, the content is sometimes left aside, asit were, and people seem to take into account many other aspects, e.g.,the position of the speaker, the circumstances of the utterance, and soon. The limiting case here is that of statements expressed in some rituallanguage people barely understand, or do not understand at all, yetjudge 'true';

(iii) people who say that a given statement is ' true' sometimes do not behaveat all in accordance with their alleged conviction. The healer evoked inchapter 2 mentioned ant-hills as an index of witchcraft activities; butpeople who take this to be true are not especially distressed by thepresence of an ant-hill in their backyard, and in any case they would notmake too much of such 'hints' in everyday life;

(iv) people seem to take as ' true' many counter-intuitive statements whichconflict with what they admit in other contexts, notably in theireveryday life. They do not seem at all worried by these apparentcontradictions, and this makes it difficult to say that they take bothkinds of statements as 'true'.

The list is neither complete nor very sophisticated in its formulation, butit should give an idea of the difficulties encountered. Obviously, theseproblems are connected; as we will see, most anthropological reflections onthe question aim at solving them all in one shot, as it were, by focusing onone of these questions and giving it a simple and plausible solution. In thefollowing pages I will examine some such ' solutions', and show that they arenot entirely satisfactory, mainly because the problem itself is not identifiedprecisely enough.

In a first approximation, the problem is that people involved in traditionalinteraction seem to apply the predicates 'true' and 'false' in a way thatwould not 'pass muster' in the anthropologist's culture. They apply thepredicates to things which, apparently, just cannot be true, and for reasonswhich do not seem conclusive at all. Now there are three obvious ways ofinterpreting this, one may assume:

(i) that the people concerned are simply illogical, and do not really have therelevant cognitive equipment;

(ii) that there is a linguistic misunderstanding; the informants either do notreally think the statements are ' true' or have a special concept of' truth'which is adapted to such statements;

(iii) that the odd statements are stored in a special compartment, as it were,so that ordinary criteria do not apply to them.

47


These three types of interpretation can be found in the anthropologicalliterature, sometimes in combination. The first approach is cited here for thesake of exhaustiveness; extravagant claims about people having no notion ofcontradiction or logical entailment are not taken seriously or even discussedanymore. The second, 'linguistic' interpretation is implicit in manyanthropological descriptions, but it is seldom justified in theoretical terms.Although it should be sufficient on its own, it is often found in combinationwith the third approach, which constitutes the most widespread interpret-ation. The idea is that in traditional contexts people have a special way oforganising the information conveyed, and more precisely, of relating beliefand experience. This is the starting point of most 'theories of belief inanthropology, which I will examine presently.

Although these approaches are at the centre of well-known, classicalcontroversies, it may be of help to reconsider them in the perspective outlinedat the end of chapter 2. Only some discourse registers, i.e., only certainspeakers in certain circumstances, are supposed to say the truth aboutspecific topics. Why and how are these speakers, contexts and registerssingled out? As we will see, anthropological ideas on truth only providepart of the answer, and some of them provide brilliant solutions for irrelevantquestions. Surprisingly, a satisfactory account of traditional 'truth-making'mechanisms will not necessarily focus on the status of' traditional belief andon traditional 'conceptions of truth', but on much more limited aspects ofcognitive processes. In order to show this, we must briefly examine thelimitations of both types of interpretation.

Truth-terms and truth predicates

Let me first consider the idea that different cultures have different'concepts' of truth. The English terms 'true' and 'false' have a kind oflogical transparency or obviousness that cannot always be found in thelanguages of the societies concerned. In many languages, the terms used toconvey the ideas of truth and falsity are not that simple: there can be severalterms with different implications, and in most cases the words have other,non-logical meanings. The Fang for instance would make a distinctionbetween meduk and abele, which could be glossed as 'complicated, intricate,obscure, false' as opposed to 'simple, clear, true'. There is a translationproblem here, as the equivalence of such terms and our ' false' and ' true' isnot self-evident. The Fang notion seems more complex than an abstractlogical predicate; it seems to convey some implicit description of what makesthe difference between true and false statements: in the Fang case, thereseems to be some connection between a statement being judged ' simple' andits being true. This is by no means exceptional; in most cultures, the wordsused to describe the truth or falsity of an utterance happen to be conventional

48

Criteria of truth

metaphors. The most telling example might be the ancient Greek aletheia,which could be glossed as 'no-forgetting', 'not-forgettable' or 'attention-demanding'.1

A crucial anthropological problem is to account for the strangeness ofcertain traditional beliefs. Given the apparent absence of strictly logicalterms, it is obviously tempting to correlate these two phenomena, and toimagine that people have ' strange' ideas of what is true and what is falsebecause their very concept of truth is different from that of the anthropologist.In this account, the associated or metaphorical aspects of the termsconsidered are in fact central. In his account of Chagga ideas of truth, forexample, Gutmann emphasises the links between such related concepts aslohi, ' true' and iloha, the generic name of sorcery rituals; the implication isthat true statements are seen as akin to magical incantations, i.e., bindingwords (Gutmann 1926: 703, cited by Steiner 1954: 366). We may suppose, inmuch the same way, that the Fang have a certain complex concept of' truth-as-simplicity', which cannot be captured by any English word. This idea iswhat makes the Fang judge as ' true' certain utterances which we would nottake in that way. This seems a simple explanation of both the strangeness ofthe beliefs and the absence of strict Fang equivalents for the English logicalterms.2

However, the plausibility of this idea vanishes if we examine the linguisticand cognitive processes involved. First let me make a strict distinctionbetween truth predicates and truth-terms. A truth predicate is an abstractreality, it is a logical term that applies to sentences that say, of what is, thatit is, and of what is not, that it is not.3 A truth-term is a word which(supposedly) expresses the predicate in a certain natural language. Not toconfuse logical predicates and natural language terms, let us call thepredicate T and — T; we can say that the French vrai and faux or the Italianvero and Jalso are truth-terms used to designate the concepts T and — T, inFrench and Italian respectively. Now we can rephrase the anthropologicalidea in more specific terms. The assumption is that the truth-terms of thelanguage can serve as transparent evidence of the type of concepts peoplepossess. If there is no simple term for truth predicates, it means either thatpeople do not possess this type of logical concept at all or that they have amore complex idea of the properties of utterances.

However, both conjectures are highly implausible. Take the example ofFang, in which one finds the terms I glossed as 'simple' and 'complicated'.One could infer from this that the Fang have some idea of simplicity/complicatedness which they use to evaluate people's utterances and come toa conclusion as to their being or not being accurate descriptions of states ofaffairs. It is not quite clear, however, what kind of intellectual operations thiswould imply. I am French; if I say I am French, Fang listeners find thisabele/'simple'; if I maintain I am Italian, it is meduk/'complicated'. The

49


surface complexity of the utterances is not taken into account; whatevercomplicated sophistry I use to ' prove' that I am French, people will find theutterances abele/'simple'. So the idea of'simplicity-truth' cannot be linkedto syntactic or rhetorical complexity. It cannot be reduced to the choice ofwords or phrases either; many formalised traditional statements are veryobscure to most Fang people, who nevertheless find them 'simple'. If wecarry on, trying to define what 'abele' sentences have in common, we willreach the conclusion that they share no common property at all, except that,for the listeners, they say of what is, that it is and of what is not, that it is not.In short, even if we believe that a strange truth-term expresses somethingdifferent from truth predicates, we will end up describing this 'somethingdifferent' as, precisely, the truth predicates. Therefore the hypothesis, that a'strange' truth-term implies a different concept of truth, seems untenable.

This crude thought-experiment is obviously not all there is to say about thepresence of conventional metaphors in truth-terms; as Steiner remarks, theemotional aspects beside the logical use of such terms are obviously crucialfor an anthropological account (Steiner 1954: 364). There is a relationshipbetween the metaphor and the logical use of the term, but it does not go inthe direction anthropologists would sometimes assume. To return to theFang example, people say, about utterances that describe non-existentsituations, that they are meduk/'complicated'. From the use of the term insuch contexts, they can form some idea of what false sentences are typicallylike: they consist of'complications' devised by cheaters in order to fool otherpeople. This is consonant with the idea that important false statements donot stem from errors but from conscious attempts to conceal some reality. Inany case, what is important here is that such vague ideas are never used ascriteria to decide what is true and what is false; rather, they are typicalimages built from the observation of what has been said to be true and falsein the first place.

This is why I used the notion of conventional metaphors in the descriptionof these terms. It is a very common feature of natural languages, that theorganisation of certain semantic domains can be described only by usingterms derived from another experimental domain. Intellectual arguments, forexample, are commonly described in English by using the vocabulary ofwarfare (a theory is 'attacked', one prepares a 'line of defence', and so on;see Lakoff and Johnson 1977: passim, and also Lakoff and Koveceses 1986).In much the same way, anger seems to be described in many languages interms of a (hot) liquid spilling over a container. Such usages are not literal,but they do not constitute 'live', poetic metaphors either. The use of truth-terms like 'simple' or 'clear' seems to be based on the same type of process.The important point here is that using such conventional phrases does notimply taking their implications literally. People who talk about anger'pouring over' do not believe they have a small bottle in their hearts which

50

Criteria of truth

spills over when they run out of patience. In the same way, people can findprototypical true discourse 'clearer' than typical lies, but they do notmeasure the 'simplicity' of utterances or use it as a criterion of truth.4

The mistaken inference from terms to concepts has an immediate appeal,because it seems strange that a language should have no specific word toexpress such simple predicates as 'true' and 'false'. From this, one isnaturally led to suppose that people who express themselves in that languagecannot possess the logical truth predicates. But this is a gross mis-representation of the relationship between words, conventional metaphorsand cognition.5 The fact that a concept is not expressed by a single term inthe language does not mean that speakers of that language do not possess theconcept, and this is clear with regard to colour terms as well as logical terms.6The case of truth predicates is by no means exceptional, and many languageslack those meta-linguistic terms which seem so natural to English speakers,like 'word' or 'sentence'. So the strangeness of some traditional truthscannot be explained as an effect of strange concepts of truth.

The logical status of beliefs

To most anthropologists, examining the question of traditional truthsnecessarily implies dealing with questions of belief. You can explain howcertain statements are judged true if you have a good theory of themechanisms of belief. This idea is so entrenched in the discipline that it takessome explanation to realise that, from a strictly formal point of view, thingsshould go the other way round. Having a set of 'beliefs' (in theanthropological sense of that term) is representing a set of propositions andrepresenting that they are true. Having the belief that ant-hills are an indexof witchcraft implies representing (i) that ant-hills are an index of witchcraftand (ii) that the former proposition is true. But no-one can represent bothwithout having some idea of what it is for a proposition to be true. Soquestions of truth are logically more fundamental than questions ofcommitment or adherence, the usual object of theories of belief. Inanthropology, however, things are presented in the other direction.Adherence or commitment is the prime mover, as it were, and truth-ascription one of its consequences. For instance, a witch-doctor says thatant-hills mean witchcraft; people have good reason to trust the speaker; theyadhere to the statements he makes; they consequently judge them true.Stated in those vague terms, the proposition is certainly true. In the followingpages and chapters I will try to show that it is also extremely ambiguous, andthat it is well worth trying to spell out the logical steps in the process ofpersuasion and adherence. Let me now proceed to anthropological ideas ontraditional belief.7

The starting point in most theories of belief is the strangeness of some

51


exotic beliefs, and the main problem is that of the apparent irrationality ofpeople who hold them. How can sensible people really claim, for example,that you can get rid of diseases by reciting some appropriate spell? Thequestion would be easily solved if people were just blatantly illogical orirrational, but they are not, and in their everyday life display a rather cool,rational way of treating problems and evaluating utterances. So the problemis about the processes whereby usual, everyday rationality can be combinedwith odd claims, especially those made in traditional contexts. This is themost interesting aspect of the old controversy known as 'the rationalitydebate' (see Wilson ed. 1970, Hollis and Lukes eds. 1982).

Inasmuch as the problem is a cognitive one, a problem of informationprocessing, an obvious solution is to posit that everyday beliefs andtraditional claims are not stored or processed in the same way. This is whytheories of belief generally focus on the special status of traditional, formalor ritual statements. The distinction can be justified in terms of cognitivesalience, although the authors seldom mention that aspect. Ritualised orformalised statements stand out as especially relevant, profound andpsychologically salient, and cannot be represented in the same way as thoseordinary beliefs which are not really said to be true, simply becauseeverybody takes them for granted. Take for instance two statements made byFang speakers during my stay:

(a) The sun causes the heat during the day(b) Ghosts dwell near dark puddles in the forest

Both were held true, so one may think they are both ' beliefs' in the ordinarysense. But the difference is obvious in people's reactions to the denial of suchstatements. Denying the effects of the sun is denying the obvious, since onlysmall children could be unaware of such facts of life. People always took mydenial (rightly) as a joke or a figure of speech. Denying the truth of (b), onthe other hand, is a more complex matter. I heard it contested by some Fangspeakers, but the audience who had previously judged the statement to betrue did not really react much. In fact, stating and denying (b) have onecommon presupposition, namely that the listeners are invited to represent thespeaker as someone who has a specific knowledge of questions of witchcraftand ancestors. I will return to this special feature presently. Suffice it to saythat the interpretation of denials, like cognitive salience, memorability, etc.,make obvious the distinction between everyday, common-sense beliefs andwhat is asserted in traditional situations.

The aim of theories of belief is to account for this difference in plausiblepsychological terms. Having such an account is necessary because manytraditional beliefs happen to be incompatible with what is held by commonsense. A usual solution is to state that traditional beliefs are stored in aspecific box, as it were, and are thus shielded from confrontation with

52

Criteria of truth

common sense. Their representation, treatment, memorisation involvespecial mechanisms. An example of such an interpretation is M. SouthwolcTsdistinction between 'religious' and 'factual' beliefs (1979: 63Iff). Factualbeliefs are just beliefs in the ordinary, philosophical sense, i.e., propositionsheld to be true, without necessarily being represented. Most people can besaid to believe that salt is white and water usually liquid, even if it is notexplicitly asserted. Moreover, people also have beliefs that have never beenrepresented at all; all my readers believe that Cleopatra did not use a word-processor, although they have never represented it.' Religious' beliefs on theother hand (which are not necessarily limited to the ordinary extension of theterm 'religion') must be represented. If we maintain that the Fang believe inthe efficacy of witchcraft, we imply that they do have some explicitrepresentations about witchcraft being efficient, be that a set of general ideasor some memories of specific rituals. Moreover, religious beliefs are oftenexplicitly asserted. So we are not dealing with the same type of processes atall. The reasonings whereby an explicit statement is judged true are not thesame as those involved in accepting, e.g., that unsupported objects tend tofall downwards rather than fly in the air.

This is certainly true but insufficient on two accounts. First, there areindefinitely many statements explicitly asserted, which have nothing to dowith 'religion', even widely construed. Even if explicit representation isnecessary, it is therefore not a sufficient condition for a belief to be treatedas 'religious'. We need more specifications. Southwold's idea, however, hasone interesting consequence. By insisting on the explicit character of thosebeliefs as a relevant property, it suggests that what is specific about, e.g.,traditional ones may well be the way they are asserted, a point I will examinein the following chapters, from various viewpoints. The second problem withSouthwold's distinction is that it does not specify the different processes of,e.g., representation and reasoning, applied to factual and religious beliefsrespectively.

This is precisely the starting point of D. Sperber's more sophisticatedaccount of the distinction (1982: 171-7). For Sperber, the main differenceinvolved here is not so much a question of content as one of logical status.Having a belief implies representing a certain sentence, which expresses the' object' of the belief. It is possible to distinguish between sentences which aresimply represented as such (e.g. 'salt is white', 'fishes have no legs') andsentences embedded in a commentary of the form 'it is true that " p " ' (e.g. 'itis true that" inflation is bad for the economy " ' , ' it is true that" ant-hills meanwitchcraft'", etc.). Factual beliefs are the former and representational beliefsthe latter type of representations. The point is that most traditional truths arerepresented in a 'representational' format, e.g., of the form 'the properinterpretation of " p " is true'.

There are some obvious advantages to this proposal, if we consider the

53


problems listed at the beginning of this chapter. For one thing, as Sperberhimself points out, a sentence can be embedded in this way without beingtotally interpreted, and that is precisely what happens, for example, withpopular beliefs about specialised subjects like economics: most people wouldcertainly admit that the sentence 'inflation is bad' expresses some truth,without really being able to explain in what way inflation is bad. All theyknow is that most experts agree on this point, so it must be true; they ofcourse have some vague images as well, about what inflation is like (Sperber:1982). This may explain how people can assert that a certain statement istrue, without feeling committed to its logical consequences. They can admitthat 'ant-hills mean witchcraft' is true and not bother to notice ant-hillsaround them; what they are committed to, in fact, is that some truth isexpressed by the proper interpretation of that sentence, although they are notsure what the proper interpretation is (e.g., may certain ant-hills meanwitchcraft in certain contexts? etc.).

Southwold's and Sperber's distinctions make it possible to eliminate someapparent problems with traditional truths. Yet they cannot constitute ageneral solution, without implying some weird claims about intellectualprocesses. This in fact is a general defect of the hypotheses proposed in the'rationality debate' and the main reason why this debate always seemssomewhat irrelevant to most anthropologists. The main problem we startedwith was to explain why and how people find true the statements expressedby certain speakers in certain contexts. This is the basic problem mostanthropologists have to solve; its formulation, however, is ambiguous, andI think anthropological theories of belief constantly deal with only one,irrelevant aspect of it. What the 'rationality' hypotheses explain, when theyare plausible, is how people can hold certain statements true and remainreasonable. That is, people who claim that a ghost can be in two places atonce and lift mountains nonetheless lead normal lives, in which they considerthat beings have only one location and limited physical capacities. The basicsolution is to have the traditional beliefs cordoned off, as it were, so that theydo not contaminate everyday cognition. Now, even if these hypotheses aretrue, they are insufficient to solve another problem, which in fact is the realanthropological one, namely why and how people find these statements trueand not others.

Anthropological theories have consistently focused on the differencebetween traditional and everyday beliefs. In the account of traditional truths,however, this point is only of marginal interest. What should be described iswhat makes certain non-everyday beliefs more salient and memorable thanothers. This is why I have restricted myself to a minimal discussion of theanthropological theories of belief; insofar as their aim is to refine the logicaldistinctions between various kinds of belief, they are always missing thecrucial anthropological problem, and at best provide necessary yetinsufficient conditions for traditional statements.

54

Criteria of truth

The rationality debate focused on the idea of' strangeness' on the fact thatsome people seem convinced by seemingly irrational statements. Thereforemany participants in that debate believed that once the problem ofstrangeness is 'solved', there is nothing more to say about traditional truths.This has the strange implication, that in order to explain why people holdsome proposition true, it is enough to show how its refutation is blocked. Therationality debate in its entirety is an illustration of this idea. The recurrentexamples are situations like, for example, the Nuer saying that 'twins arebirds'. Now the Nuer also know that twins are not birds; they cannot fly,they have no feathers and lay no eggs, etc. And the problem is to describeNuer mental processes in a way that makes it possible to hold both types ofbelief and stay reasonable. It should be obvious, however, that this still doesnot explain why Nuer people believed twins to be birds in the first place.What makes the idea of birdlike twins so salient?

Pushed to their logical consequences, theories of belief would suggest thattwins are said to be birdlike simply because (i)' that's the tradition' and (ii)the idea is shielded from empirical refutation. In other words, people aregiven a catalogue of beliefs with no intrinsic justification, and whoserefutation happens to be blocked. They therefore believe them. This ofcourse does not hold. There are indefinitely many beliefs, whose refutationis blocked or impossible, and which no-one pays any attention to.Furthermore, there are indefinitely many beliefs in the current repertoire ofany human group, which will be forgotten, left aside or transformed. Somestatements, however, will become psychologically salient, will be memorisedand repeated in subsequent occurrences. Why are they treated differently?This is a crucial question for the study of tradition, a question about whichtheories of belief have nothing to say.

Conceptions and criteria

Obviously, the strategies examined here do not provide a satisfactoryaccount of the problems of traditional truths. The approach which considersthe logical status or the overall organisation of beliefs describes only part ofthe cognitive processes involved; it does not account for the particularsalience of traditional truths. The interpretation, which stresses the influenceof linguistic factors, generates some untenable hypotheses unless it issubstantially qualified. In fact both approaches have a common defect, inthat they are far too general to give any empirically significant explanationof traditional truths.

In both conceptions it is taken for granted that the problem requires ageneral solution. The way statements are judged true in traditional contextscan be explained by resorting to general hypotheses, about the waypropositions are stored or the type of logical predicates available in thelanguage. The problem with such solutions is that they are always either too

55


weak or too powerful. Theories of belief, for instance, when they areplausible, are much too powerful. They suggest that people, if given anexplicit catalogue of statements with blocked refutation or' representational'status, will believe them all. This is obviously untrue, and the theory is notfine-grained enough in its description to account for the fact that only somesuch statements are actually believed and memorised. The linguistic analysison the other hand is much too weak, in the sense that if it were true, if peoplereally had complex concepts irreducible to truth predicates, it would be verydifficult for them to hold true even the most pedestrian everyday statements.

The reason why these interpretations are unsatisfactory is that they aredirected at the wrong kind of intellectual objects. Here I must emphasise thedifference between conceptions and criteria of truth. Conceptions of truth, ina general sense, can be described as sets of general propositions about therelationships between sentences and states of affairs in general, or about themeaning of truth-terms. The interpretations described here are entirely aboutthe conceptions of truth implicit in traditional statements, and they are basedon two assumptions, namely that when people judge a given utterance true,they are implicitly, perhaps unconsciously, applying certain abstractprinciples about language and reality to the case at hand, and that theprinciples applied in traditional contexts are different from those obtainingin other cultures or contexts.

An extreme form of this idea is M. Hollis's hypothesis, that in ritualcontexts people abandon their usual conception of truth as correspondence(what is true is what describes the world the way it really is) to adopt a'coherentist' view (statements are true if they 'fit' in what one already knowsabout the world) (Hollis 1970: 223). Now this is unsatisfactory, not only forempirical reasons (people sometimes find true ritual statements whichcontradict what they usually admit), but also in principle; I will argue thatthe difference between the veracity of traditional, ritualised statements, andthat of ordinary discourse, just cannot be a question of conceptions of truthand therefore has to be explained by specific criteria of truth.

The fact that conceptions of truth, in whatever form, will give nosatisfactory answer, should be clear from the fact that such explanations, asI said above, are always either too strong or too weak. Hollis's hypothesis isno exception; if you admit that traditional statements are judged on theircoherence with other beliefs, then you make it possible for people to entertainindefinitely many beliefs, which in fact they would never hold. This recurrentproblem is in fact the consequence of a logical mistake. It is probably truethat any specific truth-ascription implies some conception of truth; but itwould be wrong to assume that the converse is true, namely that a conceptionof truth implies specific truth-ascriptions, or in other words that specifictruth-ascriptions can be deduced from a conception of truth. This is clearlyfalse. People do not judge statements on the sole basis of their (unconscious)

56

Criteria of truth

theory of truth, they have many other reasons for finding specific statementstrue or false. This is where criteria of truth become pertinent.

Criteria of trugh are specific reasons for finding an utterance true. Theyconcern the content of the utterance, its relationship to other propositionsheld to be true, the person of the speaker, an evaluation of his or herknowledge, and so on. They cannot be generated by abstract principles, theyare grounded in specific descriptions of the utterances concerned. The Fangjudgement, for instance, which decides it is true that ant-hills are signs ofwitchcraft, is not the result of a transcendental deduction or of a special useof the term 'true'. Fang listeners take the statement to be true because theyknow this specific speaker is reliable, has been initiated into anti-witchcraftrituals. Criteria are diverse and can be conflicting. Take for instance thestatement' there are three sexes in giraffes' uttered by a reputable biologist;in such a case, the evaluation of the speaker and of the content may beconflicting for a listener with a poor knowledge of biology. It is a well-knownfact of life that the judgement on many statements is thus suspended, eitherbecause the evidence is lacking or because it is conflicting.

Criteria and criteria types

Obviously, if we want to explain why people hold true certain statements andnot others, we will need to examine their criteria of truth. If we want tounderstand why certain speakers in certain contexts are supposed to speakthe truth, we must study the criteria applied to those precise contexts andspeakers. The reason why this is not recognised in theories of traditionaltruth is that the task seems impossible; since criteria are specific, concern thespecific speaker and context, they must be as many as the contexts studied.The criteria applied to Fang diviners will not be the same as those relevantfor Shilluk priests or Zande witch-doctors. Therefore the study of criteria isa question of ethnographic description, not anthropological theory. Nothinggeneral can be said about them.

This is not entirely true. Criteria are indeed specific, but this does notpreclude them from being classified in a few general types. If, for example,some people hold true whatever is said by reputed physicists, reputedbiologists and reputed sociologists, it is likely that the criteria applied todifferent contexts belong to a single criteria type, to do with reputedscientists. It must be possible to establish a taxonomy of criteria of truth. Ifwe assume that traditional truths are special in some way, we will try todescribe the common features of the criteria applied to traditional statements;in other words, our taxonomy will have a special node labelled 'traditionalcriteria type', maybe with a few appropriate sub-types, to distinguishbetween different kinds of traditional contexts, as the case may be.

It is therefore conceivable that there is a common criteria type for

57


traditional contexts in general. In fact, the 'common' anthropologicalconception of tradition includes a hypothesis about that criteria type.Unfortunately, it is a very poor hypothesis. Its main theme is what could becalled the principle of authority. We know that in traditional contexts truthis reserved to specific speakers and specific circumstances. The principle ofauthority says that there is no mystery at all in this process; in a traditionalcontext, whatever is said by certain people or whatever is said in the contextof certain ritual operations is true. The Fang believe that ant-hills are anindex of witchcraft because it is said by a witch-doctor during the appropriateritual. The criterion of truth is that whatever witch-doctors say is true; if weadd that whatever the elders and the ancestor cult specialists say is true, wehave a fairly complete list of Fang criteria. As for the general criteria type,it could be formulated like this: whatever is said by persons who hold someknowledge bequeathed by former generations is true.

It is important here not to be confused about who is making the hypothesis.We suppose that people involved in traditional interaction, when theyevaluate the statements made in specific contexts, represent those statementsas the consequence of a 'competence' that is transmitted from generation togeneration. The common conception suggests that this is in the head of thepeople concerned. As we saw in chapter 1, the common conception alsoimplies that this is the real state of affairs, that there is a real competence anda real transmission. These are two different hypotheses, one is about the waypeople interpret some utterances, the other about what really happens. Inchapter 1, I presented various arguments against the idea that things reallyhappen that way, against the idea of a real transmission of world views. Evenif my arguments were right, it might remain that the 'transmission ofcompetence' is the interpretation made by the actors involved in suchsituations. That is what we want to examine now. Is this idea of traditional'authority' really implied by people's reasoning about traditional utterances?

As a description of people's reasoning, the 'principle of authority' isobviously unsatisfactory. It has some prima facie evidence; people do saythings like 'this is true because the initiates say it'. The principle, however,is insufficient and question-begging. It is insufficient, in the sense that everyanthropologist has met sceptics and dissenters, and a description oftraditional criteria should account for such cases, which become in-comprehensible if established competence is all there is about traditionaltruth. Moreover, the hypothesis is question-begging. It assumes that peoplefind true, for instance, whatever is said by ritual specialists. But thesespecialists are recognised as legitimate, because they have certain character-istics which make them 'truthful'. To return to my Fang example, the witch-doctor's utterance (about ant-hills as indices of witchcraft) is judged truebecause the speaker is an evur-bearer, has performed the evur-shapingrituals, has undergone a specific initiation comprising some direct dealings

58

Criteria of truth

with ghosts, and so on. That Fang listeners find his utterances true ' becausehe is an established witch-doctor' is right only as a shorthand version of theinteraction. But in the process of shortening it we have left aside the crucialreasonings which are relevant to truth-ascription. In all logic, truth here isnot a function of a social position, but a consequence of the specificrepresentations people have about the processes leading to the position.

The same argument applies, mutatis mutandis, to contexts instead ofpositions. It is sometimes stated that the participants in traditional ritual findtrue whatever is said by possessed people. Again, this is just shorthand.The criterion of truth is not that possession implies truth, but that possessionis represented in such a way that it is likely to result in the production of trueutterances. If you use the shorthand version and forget that it is justshorthand, then you have the 'principle of authority', saying that whateveris said in proper ritual contexts is true. But that does not hold as a descriptionof the participants' reasonings.

All this may seem obvious. Indeed, it is obvious, and ethnographicdescriptions of particular traditional situations always give the 'longhand'version of the reasonings involved. They give all the relevant specificrepresentations which make witch-doctors or possession truthful. Inanthropological theories, however, we only find the shorthand idea thattraditional truths are judged true because of the authority of position andcontext. This happens because the longhand reasonings I have mentioned aresupposed to be specific to each culture and context, and to have nothingmuch in common except, precisely, the idea of authoritative social positionsand favourable contexts. The Fang reasonings about witch-doctors and theZande reasonings about diviners are supposed to be totally different, the onlycommon idea being that authority is vested in specific positions andoperations.

In the following chapters I will try to show that this is entirely wrong. Thecriteria of truth and the reasoning applied to Fang witchcraft, Zandedivination, or Malagasy ritual discourse are indeed specific, but they shareimportant, non-trivial characteristics beyond the mere identification ofspecific positions and operations. The 'criteria type' applied to traditionalinteraction can be described in some detail, avoiding the question-beggingclaims usually found in anthropological theories.

Criteria and situations

As I said at the beginning of this chapter, traditional truths seem problematic;in a traditional context, people do not seem to judge utterances in the sameway as in everyday conversation. The object of judgement seems different,since people can assert that a statement is true and be doubtful about itsimplications. Also, the grounds for judgement are puzzling, in that the

59


content of the utterances is often left aside. But anthropological reflectionson truth do not have any satisfactory explanations to put forward. Instead,they focus on questions like rationality and the status of belief, which are forthe most part irrelevant to the problem at hand. Such theories only providea general framework in which it is not entirely unreasonable to holdtraditional 'beliefs'. This, obviously, is insufficient and does not even beginto explain for what reasons people hold these beliefs. It is impossible toaccount for traditional truths in a satisfactory way without focusing on thetype of criteria used. Against a common anthropological assumption, I haveargued that it is in fact possible to find common properties, specific totraditional interaction, in the reasoning whereby statements produced invarious cultures and contexts are judged true.

In the description of abstract conceptions of truth, the characteristics oftraditional truths mentioned at the beginning of this chapter pose difficultproblems. In the description of the specific criteria of truth on the other hand,these characteristics are of great help. They are common features of thetruths produced in traditional contexts, and therefore constitute an excellentstarting point for our inquiry. In the following chapters I will therefore focuson what may seem the most puzzling characteristic, namely that traditionaltruths are asserted in particular situations, which require a combination ofspecific actions and specific actors to perform them. Furthermore, theexpression of truths makes it necessary to take many ritual precautions.Diviners for instance must avoid certain foods or activities, ritual specialists'prepare' themselves. 'Precautions' or 'preparations' concern both theutterances and the persons. Special forms of speech and discourse registersare used in traditional contexts. An utterance takes on a special value, byvirtue of the actions or technical operations by which it has been produced,e.g., a divination procedure, or by virtue of the special language, orvocabulary, which is used. The precautions can also focus on the very personof the speaker. Not all speakers are in the same position as regards specificdomains of reality, not all of them are in a position that allows them to uttertrue statements. In most cases such a position is the consequence of a specificinitiation, either a 'tribal' rite everyone has to undergo or a more specialisedinitiation, such as the Fang witch-doctors' apprenticeship mentioned inchapter 2.

Although such precautions are always mentioned in ethnographic studies,they are supposed to be part of those contextual factors which provide thesetting, as it were, whereas the content of the utterances is essential. Thisdescription, however, is misleading; to the people concerned, theseprecautions are essential ingredients in the production of truths. Traditionalstatements are judged true because they come from 'customised' persons, asit were, and require 'customised' speech. In order to describe the type ofcriteria of truth applied to the traditional statements, we must thereforeexamine both types of customisation.

60

Customized speech (I): truth withoutintentions

In many or most cultures, divination is considered a reliable source of truestatements about important events or states of affairs. In traditional contextsit constitutes a crucial 'truth —making' procedure. Anthropological theories,however, do not say much about the intellectual processes wherebydivinatory statements are judged true. Although there is a number of precise,sometimes illuminating hypotheses about the social contexts of divinationand 'divination-driven' decisions, the cognitive aspects of the phenomenonare relatively neglected. In this chapter I will try to put forward somehypotheses about the cognitive mechanisms which make divinatoryprocedures and the observation of omens a plausible source of truth.

Anthropological arguments: a school of red herrings

Before examining some general properties of divination procedures, let mefirst focus on some common anthropological ways of avoiding the questionof veracity. They constitute a vague pre-theoretical picture of divination,which interferes with the anthropological description but cannot be taken asa serious theoretical attempt. The common picture usually implies somegeneral claims about traditional statements, like the fact that divinatory truthis a function of authoritative positions, or that the statements made indivination are not supposed to be 'true', but simply 'relevant' or'meaningful' (see chapter 3). While the idea of truth reduced to 'meaningful'statements is simply untenable given the ethnographic data, the 'principle ofauthority', taken as an explanation of truth criteria, just begs the question.Some claims, however, concern the specific features of divinatory truth; theycan be summarised as follows:

1 The divinatory statement is manipulated. This is one of the commonestarguments. It runs as follows: although diviners try to convince theircustomers that they are just communicating the results generated by a neutralprocedure, they actually manipulate the diagnosis, typically by usingbackground knowledge about the situation at hand, the clients' wishes andthe likely outcomes of present states of affairs.

61


2 The content of the statement makes it irrefutable. The idea is thatdivinatory statements are expressed in such an ambiguous way that anysubsequent course of events would confirm them. A paradigmatic case is thePythic oracle predicting the demise of a great empire, without sayingprecisely which. Mendonsa opens his account of Sisala divination with thestatement that' the viability of divination rests on the fact that a divinatorypronouncement can never be proved wrong' (Mendonsa 1982: 109). This issometimes used in conjunction with the previous argument, although eitherof them, alone, should provide a sufficient explanation.

3 Refutation is blocked by 'secondary elaboration'. This phenomenon,brilliantly described in Evans-Pritchard's study of Zande oracles (1937) isfamiliar to all anthropologists: the empirical refutation of a particulardivinatory diagnosis only casts doubt on that particular ritual, not on thedivinatory technique in general. People find special reasons to explain awaya particular failure, by assuming, e.g., that the diviner had eaten forbiddenfoods or that the particular ritual was not properly performed, so that theycarry on relying on divination and oracles in general.

These claims can be found in various combinations, either explicit orimplicit, in many anthropological monographs and theoretical essays.Whatever the presentation, they do not constitute a satisfactory account ofthe veracity of divination. Each of the arguments is insufficient, in that manycounter-examples can be found, so that they are often used together in ad hoccombinations. In order to go beyond this composite, and vaguely inconsistentview of divination, it may be of help to examine the flaws of each argument.

It is somewhat ironical, that anthropologists dealing with the veracity ofdivination manage to avoid the question by resorting to the standardarguments of enemies of divination or omens. Thus, the idea of amanipulation of diagnosis, based on background knowledge, can be found inmany pre-anthropological writings. The idea that diviners try to keep theircustomers happy, by telling them what they want to hear, can be found inCicero's De Divinatione. The idea of background knowledge as the real basisof relevant diagnoses is so pervasive in the anthropological literature that itcontaminates outsiders. K. Thomas for instance states, in his account ofclassical magic and religion, that English 'diviners...like their Africancounterparts, maintained their prestige by a combination of fraud and goodpsychology' (1971: 289, see also 50ff.). This argument is clearly insufficient.Not all divination rituals make such manipulation possible; on the contrary,many procedures are organised as series of yes/no questions which leave nointerpretative leeway to the diviner. Moreover, if diviners were experts atmaking judicious statements, the problem of empirical refutation shouldnever arise. But it does arise, so the argument is at best insufficient, and inany case would not explain why the procedure should be taken as anindispensable element.

62

Customised speech (I)

The second argument is based on the alleged difficulty of refutingdivinatory statements, because of their intrinsic vagueness. Again, thisanthropological hypothesis about divination is the modern version of aclassical argument against divination.1 The argument is based on thehypothesis that, in certain circumstances, subjects hold a statement true justbecause its refutation is blocked. This, as I pointed out in chapter 3, isobviously insufficient, and certainly does not explain why a statement isconsidered plausible in the first place. In any culture, there are many contextswhere people make claims whose form or content puts them beyondrefutation. Not all such claims are considered convincing, so that adescription of divination simply cannot avoid the question, of why divinationshould be convincing. Moreover, many divinatory statements are specificand predictive enough to be refuted by subsequent events. This is where the' secondary elaboration' argument comes into play. This idea, although verypopular with anthropologists, is not really satisfactory either. True, peopledo resort to such reasonings to protect the basic tenets of their beliefs,including their faith in divination, but we are left again with the crucialquestion, why should it be protected in the first place? There must be somereasons why divination procedures appear convincing, and therefore worthprotecting, which the notion of 'secondary elaboration' does not explain.

To summarise, these statements about divination just beg the crucialquestion of the veracity of divinatory diagnoses.2 Moreover, this common(implicit) conception goes against what the clients themselves say aboutdivination. The diviners' clients consider that the statements are true becausethey have been generated by the technique. They must have somerepresentations which make it plausible that utterances brought about in thisway are true. While they insist that a divinatory statement is different fromordinary discourse, in that it is more certainly true, the anthropologicalconception implies that this difference is just a matter of decorum or cunning.If this is the case, the universal existence of divinatory procedures isinexplicable. In the following pages I will try to show that it is possible to putforward a theory of divination that goes beyond these question-beggingideas.

Elements and types of divination

Divination rituals are supposed to generate true statements. This implies thatthe participants in the rituals must represent a specific criterion of truth, whichapplies to statements generated in this way. Anthropological models tend tosuggest that the veracity of divination is just a 'cultural axiom', which isadmitted without any justification, but this is an ambiguous claim. True,people who participate in divination rituals seldom bother to elaborate anabstract and explicit 'theory' of divination and truth. But most criteria of

63


truth are implicit, in everyday life as well as in science or tradition. As Imentioned in chapter 3, criteria of truth are not necessarily generated byabstract theories of language and reality; they are specific empiricalhypotheses, which concern a certain domain of reality, certain ways ofmaking statements and inferences, and so on. These hypotheses must focuson three main elements:(i) the situation, i.e., the undefined state of affairs which is supposed to be

the cause of the problem at hand. People who consult a diviner musthave the idea that there is 'something the matter'. In a domain to whichthey have little or no access, there is a definite state of affairs which haseffects in their lives. Although people generally have no specificrepresentation of what that state of affairs is, or only have conjectures(otherwise they would not need a diviner), they know that the state ofaffairs exists and has some effects;

(ii) the procedure, the set of operations which make divinatory situationsdifferent from ordinary situations of communication. It is important tonote here, against the widespread ideas mentioned above, that theprocedures involved are always taken as what actually produce thestatements. The inspired mediums attribute the message they com-municate to some external entity; the statements would not have beenthe same if they had not been possessed or otherwise influenced by thatentity. In the same way, the result of a technical operation, e.g., shakingdice, is conceived as the foundation of the divinatory statements,

(iii) the diagnosis, i.e., the description of the 'situation' at hand, which isachieved through the use of the procedure.3

Actors cannot participate in divination rituals without entertaining somerepresentations about these three elements and their combination. In whatfollows I will argue that if the elements are represented as having certainspecific connections, then the situation as a whole can be considered morelikely to generate true statements than other discourse situations. In order toshow this, we must examine some (non question-begging) anthropologicalideas on people's representations of the divinatory activity. I will focus ontwo fundamental ideas which can be found in anthropological sources,sometimes in combination. Both contain a grain of truth and point at someimportant aspects of the rituals; unfortunately, we will see that they are notsufficient as a description of the criterion of truth used by clients in theinterpretation of divination.

Before describing these models, however, it is necessary to give a briefsurvey of the types of institutions to which they are supposed to apply. Theterm 'divination' covers various types of rituals; and the anthropologicalmodels are generally inspired by divinatory procedures of a given type, thusneglecting some obvious features of the other types. To give but a brief

64


survey of the varieties of divination, we must distinguish between two broadtypes of procedures, and also contrast these 'technical' forms of divinationwith the observation of omens and portents.

First, most authors agree on a distinction between two types of divinationfounded on the type of technique used. Divination can be performed, eitherby putting a specialist in an appropriate state of ' inspiration' (notably incases of possession), or by using specific instruments: cards, cowries, dice,sacrificed animals, etc. In the latter case the interpretation of mantic signs canbe given in advance, as e.g., in the Mesopotamian 'catalogues' which listthousands of omens and portents and their interpretation. The distinctionbetween 'inspired' and 'constrained' divinations is always present in theanthropological literature, although the exact terms may vary: Reynolds(1963: 118) calls them 'mental' and 'mechanical', and Bottero (1987: 58)'inspired' versus 'deductive' procedures. There is in fact a continuumbetween totally constrained techniques in which the interpretative work isready-made, as it were, and pure cases of possession. Variably 'constrained'types of divination, with different purposes and cognitive effects, can coexistin certain cultures (see Bauer and Hinnant 1980: passim). Like mostanalytical distinctions used in anthropology, this one refers more to idealtypes than to real differences, and most cases fall somewhere between thesetypes, as Zeitlyn points out (1987: 5ff.).

Another distinction is commonly made, between divination as a technicaloperation and the observation of natural omens. The latter are special in thatthey are available in the natural environment; they do not require anyintentional manipulation. While dice are thrown and cards drawn from apack, the flight of birds, the shapes of clouds and other such portents are notbrought about by a human agent.4 Again, however, the contrast holds onlybetween ideal types; in actual rituals there is a considerable overlap betweenomens and divinatory signs. The decision to select certain phenomena, orcertain aspects of these, as ominous, e.g., the flight of certain birds at acertain time, is clearly close to a technical operation which can only generatecertain formal combinations.

However blurred in actual ritual practice, these distinctions are relevant inthat they provide the starting point for anthropological interpretations of theclients' expectations and intellectual processes as regards the divinatorydiagnoses. Here I will mainly focus on two types of interpretation, which Iwill call 'intentional' and 'semiotic' respectively.

Divination, messages and * hidden9 speakers

In a widespread interpretation, divination is described as a ritual aimed atmaking explicit some messages coming from ancestors, gods or spirits. Thisis how Loewe and Blacker define it, at the beginning of their collection on

65


divination in 'great traditions' (1981: 4). The description seems naturalenough in cases of 'inspired' or 'emotive' divination, close to possession,and it is no surprise to find it explicitly defended in ancient sources. After all,the prototypical Pythic oracles were primarily conceived as indirect messagesfrom the god.5 'Inspired' divination is generally described as a ritual whoseaim is to receive explanations or orders from transcendent entities. Even incases of'constrained' or 'mechanical' procedures, this description is oftenclose to the accounts given by the participants. In the extremely variedAfrican systems it is common to interpret the diviner's task as one oftranslation; the bush-spirits or the ancestors are supposed to ' talk' throughgeomantic patterns or the behaviour of oracular chickens. This particularinterpretation has influenced anthropologists mainly because it is what thepractitioners often express, when they try to rationalise or justify theirpractice.

This interpretation, however, is not altogether satisfactory, because itrelies on an ambiguous description. I would argue that the 'intentional'model points at important properties of the divination contexts butmisinterprets them. In order to understand these features, it is necessary toleave aside, for a while, the alleged communication with transcendent entitiesand focus on the actual communication situation that is set up during adivinatory seance. Obviously, making a diagnosis through a divinatoryprocedure is not equivalent to making a common assertion. According toZempleni (1986: passim), it should be seen as a special type of 'speech act',in which a specific action is supposed to result in and guarantee an utterance,in that case an assertion; what a theory should describe is the formalconditions of this original speech act. Divination is a form of communicationbetween diviner and clients in which the pragmatic aspects of commonassertions are modified. In normal conversation, literal assertions areinterpreted as implying that the speaker is commited to the truth of what isexpressed by the sentence; asserting 'giraffes live in Africa' implies beingcommitted to this idea of their location. In divinatory utterances, however,there seems to be a dissociation between the person who actually makes thestatement (the diviner) and the person(s) or agencies responsible for thestatement. The diviners speak, yet it is not they who are committed to thetruth of their words (Zempleni 1986). Among the Tallensi, for instance, thediviner and the consulter hold a wand and move it above a divinatorymaterial consisting of several stones. The answer to each specific question isgiven by the direction the sticks point to. As Fortes points out, the Tallensi'insist that neither the consulter nor the diviner knows what is going to bethe outcome of the divination session. They insist that it is the ancestors whomove the stick and put the words in their mouth' (Fortes 1966: 419).

This is a very general property of divinations. Diviners take great pains toconvince their clients that their own representation of the situation concerned

66


is immaterial, and that their task is to translate what comes up in the stonepatterns or the aspect of the entrails; in the same way, they commonlyattribute their mistakes to the ancestors or spirits, who are said to have liedto the expert or deceived him. Now, if the diviner himself is not taken as theone who originates the message, then it seems plausible to surmise that themessage is attributed to other, transcendent speakers, e.g., gods, spirits orancestors. All this seems to suggest that the 'intentional' interpretation isright, and that the participants do represent a 'hidden speaker' as the originof the diagnosis.

This, however, is where the mistake lies. By putting the actual, humanspeaker between brackets, as it were, the ritual does not necessarily implythat another, hidden speaker is represented. In many societies, divinatorystatements can be generated in the absence of definite hypotheses about theirorigin. Some forms of divination dispense with the idea of 'another world'or 'other entities' altogether. To give but one example, French Bocagewitchcraft, as described in Favret-Saada's monograph (1980), is anything butintellectually fanciful. There are no refined cosmological ideas, no complexnotion of the various 'forces' involved, no theory of how witches manage tobring about their victim's demise. The victims are just faced with aninexplicable succession of untoward events, until some well-wisher suggeststhat recurrent misfortune cannot be attributed to bad luck, that there mustbe 'something more' to the situation. The experts involved in theidentification of the problem and the unbewitching rituals base theirdiagnoses on a fairly common interpretation of the series of cards drawnfrom a pack (Favret-Saada and Contreras 1984), which makes it possibleboth to identify the problem as a witchcraft situation and to give someindications as to the identity of the witch. The diviners are eager to convincetheir clients that their technique involves no 'mystical' aspect at all: thediagnosis is 'just there', transparently expressed by the succession of cardsdrawn. Neither the clients nor the diviners ever wonder through what processor agency a conflicting human situation can be thus revealed by a pack ofcards. There is no mention of any mediating entity, and no-one seemsinterested in pursuing the idea. Obviously, such intellectual concerns paleinto insignificance, when compared with the strong emotive aspects of thesituation; for witchcraft is always conceived as a matter of life and death.During the long series of unbewitching rituals that follow, the same principleholds; clients just' observe' that certain rituals work and others do not, andthat's that.

The 'intentional' paradigm is therefore insufficient. Diviners seem eager todismiss the idea that the diagnoses could depend upon their own intentions.Dismissing this, however, does not necessarily make room, as it were, foranother, imaginary speaker. In spite of this error, the 'intentional' model willallow us to go further in the description of the discourse situation. It implies

67


that the connection between situation, procedure and diagnosis, asrepresented by the participants, excludes the mediation of the actualspeaker's communicative intentions. In the following section I will turn toanother interpretation, which is related to this idea and suggests that theactors represent some form of direct connection between situation anddiagnosis.

Signs and diagnosis: semiotic links

In the semiotic understanding of divination, people who consult diviners,and diviners themselves, are supposed to conceive of their activity as theobservation and deciphering of signs. In this conception, the signs are eithernatural, in the case of omens and portents, or artificial, i.e., produced by adivinatory technique, and they stand for the actual state of affairs the clientwants to know about. The divinatory activity consists in deciphering thecorrect meaning to these signs, and the relationship between signifiers(divinatory elements) and signifieds may be more or less constrained,according to the particular features of the system considered.

Not surprisingly, this hypothesis is often based on the idea that thereligious activity of diviners, on the one hand, and the more pedestriancapacity of predicting natural occurrences on the other are just varieties ofthe same interpretive process. Divination is conceived of as an extension, totranscendent domains, of the ordinary observation of coincidences orconjunctions in natural phenomena. A good example of this reasoning isQuintus's defense of divination, in the first part of Cicero's De Divinatione.According to Quintus, the prediction of future occurrences is in fact a veryordinary activity, a skill derived from the accumulation of observation.Experienced sailors can predict the occurrence of a storm from theobservation of clouds, winds and waves. In an even more pedestrian way,everyone can predict from the shape of a seed, what the shape of the grownplant will be like. Divination, at least in some of its practices, is just anextension of this activity (Div. I.vii.l2ff.).

The idea that diviners and clients are involved in the deciphering of naturalor artificial' signs' for unobservable situations, seems convincing for severalreasons. First of all, it is the natural idiom of many forms of inductivedivination, and seems to give a good account of the fact that divinatory signsare often grouped in a system. In all systems of'mechanical' divination, themantic material consists of randomising procedures that give certainstructured answers. For instance, the geomantic signs are grouped inestablished patterns, which combine in larger patterns, and so on. This typeof syntactic organisation makes them look very much like any abstract4 language' generated by simple rules, and this I suppose is why many authorsfind the analogy with other semiotic systems illuminating.6

68


There are some limitations, however, to this semiotic interpretation. First,while it seems to fit cases of constrained or mechanical divination, it is arather inadequate interpretation of inspired divination. Divinatory diagnoseswhich are given in the context of trance or possession do not usually displaya system of formal signs. Also, and more importantly, the idea of signs isextremely ambiguous, and just begs the question of the fundamentalrelationship established between some state of affairs the clients want toknow about, and the diagnoses themselves. In order to show why this isimportant, we must return to very general properties of divination ritual.

The semiotic model does not describe the link between the situation andthe diagnosis in a satisfactory way. The idea that people represent thediagnosis as a series of' signs' is either trivial or false. It may be taken toimply that the participants represent the relationship between diagnosis andsituation as a relationship between signs and their referent. In thisinterpretation, the idea is trivially true. The diagnosis, after all, is supposedto be a description of the situation at hand. But giving a description ofsomething always consists in producing some signs or tokens that stand forthe thing described. If, on the other hand, the semiotic interpretationsuggests that divination provides a systematic equivalence between certainsemiotic features and their interpretation, so that there is a culturaldivinatory code, then it is clearly false in most cases. True, some societieshave developed such catalogues of divinatory signs and portents; but this isnot universal, far from it.7

The model therefore cannot solve the question at hand, namely why the' signs' produced by divinatory procedures are supposed to be more reliablethan other, ordinary 'signs'. As we will see in the next sections, there is animportant truth in the idea that the solution (the identification of thecriterion of truth) lies in people's representations of the link betweensituation and diagnosis. The merit of the semiotic interpretation is that it laysstress on the fact that this link is represented as a direct connection; as itstands, however, the interpretation is not sufficiently accurate and general.

To sum up, both general ideas on the actors' representations seeminsufficient. Both models rely on the idea that divination rituals arerepresented as communication procedures. As I will now try to show, thereare some limitations to this idea, which cannot really explain the intellectualmechanisms whereby divination is judged truthful. This question can besolved only if we have a precise description of the connection established, inthe participants' minds, between a certain state of affairs and its divinatorydescription.

69


Guarantees of veracity

It may be of help here to focus on the specific guarantees of truth which aregiven by divinatory procedures and on the factors that may disrupt the ritual.In what follows I will examine some examples, and try to outline adescription of divination which takes such aspects into account.

The first example is a fairly straightforward case of 'constrained' or'mechanical' procedure, taken from the description of the Mundang systemof divination by A. Zempleni and A. Adler (1972). The Mundang consult thediviners in order to find out the position of the cox sinri (earth-spirits) andmozumri (ancestors) as regards crucial questions, notably about politicalmatters. This is made manifest through the patterns formed by stones throwndown on the ground. These patterns are themselves integrated into largerstereotypical combinations, which give definite yes/no answers to the client'squestions. The different aspects of the client's situation are represented byseveral circles of stones in front of the diviner. The stones stand for the partsof the client's body, the members of his or her family and household, thewalls of the house, the hours of the day, the seasons, and so on. Thedivinatory session consists of a long series of 'deals', each of which issupposed to describe the state of one of these elements. More than a hundredspecific questions are thus given a definite positive or negative answer, andthe combination of all these minute diagnoses constitutes the divinatorydescription of the situation at hand.

The order of the questions on the check-list never changes; the first four'deals' are crucial, since they concern the 'state of veracity' of the divinatorysession itself. If the results are positive, then the session as a whole is goingto yield a true description of the situation; when they are negative, the diviner' shuffles' the stones and deals again, or decides to bring the session to an end.The results of the first four deals guarantee the truth of what is to follow. AsZempleni and Adler remark, 'the diviner is given an opportunity to castdoubt upon his own activity... The statements can be biased by the veryancestors and spirits whose intentions are being examined' (1972: 100, mytranslation). In other words, the Mundang rituals state, not only that thesituation is such and such, but also that the description itself is reliable.

The possibility of checking divination by another divination is in factwidespread, if not universal. We can find a similar procedure in thetraditional divination of the Sisala of Northern Ghana (Mendonsa 1982).The diviner (vugura) is typically consulted before and after funerals, first tomake sure that none of the mourners had been involved in witchcraft ritualsagainst the deceased, then to find out what the actual cause of death was. Butdiviners are also consulted about all kinds of important matters, notablyabout relationships with the ancestors. Divination gives access to the hiddenworld calledfafa\ it is 'the only legitimate mechanism whereby the Sisala can

70


receive feed-back from the ancestors' (Mendonsa 1982: 191). The Sisalareadily admit that divination is not altogether reliable. The fallibility of theprocedures is attributed to several causes. Diviners may just invent stories toplease their client; also, the ancestors may be angry with a diviner and makehim lose his ability; and it is generally accepted that not all diviners areequally insightful (Mendonsa 1982: 190; see also Mendonsa 1976).

An institutional procedure, called dachevung, makes it possible to verifydivinatory statements. Whenever a client wants to check the veracity of adiviner's assertions, he goes to another diviner and asks for a dachevung. Thisis a shortened version of the traditional ritual. The client draws several marksin the dust, some of which stand for the first diviner's assertions. The newdiviner, who is not told the meaning of the marks, points his stick towardssome of them. If he happens to choose the 'right' marks, those which standfor the diviner's statements, then these statements are confirmed.

Similar phenomena are reported by Fortes among the Tallensi, and byZeitlyn in his study of Mambila divination in Cameroon. The Tallensiconceive the failure of divinatory diagnoses as caused by the ancestors. Atthe beginning of the sessions there is a small ordeal to decide whether theancestors are willing to talk. In one such case observed by Fortes, theancestors are against talking because the diviner has not made the sacrificethey wanted (1966:418). A similar point can be made about the CameroonianMambila spider divination, as described by Zeitlyn (1988). The techniqueinvolves a spider, which is left under an upturned pot. Trying to escape, thespider moves a series of divinatory 'cards' or leaves, whose subsequentpositions give a yes/no answer to the question put by the diviner. Zeitlynreports that' the pots may also be asked whether any witchcraft is attemptingto interfere with them' (1988: 13).

There seems to be a circular aspect to these divinatory checks ondivination, which consist in evaluating a procedure by resorting to that sameprocedure. This however does not seem to bother either clients or diviners.As Zeitlyn points out, ' this is the only place where a Western logician canquarrel with their practice, but I did not succeed in pointing out thefallacy....No one I spoke to saw the problem as I did' (1988).

These examples are complex cases of'mechanical' divination. Part of thecomplication comes from the apparent circularity of the guarantee. Oneshould not, however, be blinded by the circularity to such an extent that onedoes not notice the important property of divinatory diagnoses that itillustrates. Here I will try to show that such examples are remarkable, notbecause of the circularity, but because they display some very generalcharacteristics of divination which are usually obscured. Such circular checksmay not be universal, but they provide a limiting-case, which I think may beof help in the description of the relationship established, in the participants'minds, between the state of affairs described and its divinatory description.

71


In other words, they constitute cases, in which some properties of divinatorydiagnoses are particularly salient.

Situations and their description: the causal link

As we have seen, ancestors in the Sisala and Mundang examples are both thesubject matter of divination, which the assertions are supposed to describe,and the entities that make divination possible. I will now try to make thispoint more general, and examine its consequences as concerns the link,established in the clients' minds, between the state of affairs described and thedescription itself. My hypothesis is that what the diviner states, with the helpof various technical procedures, is considered true insofar as it can berepresented as a picture influenced by the very entities described (howeverindirect the influence). This suggests that such influences are the criterion oftrue mantic statements in general; the case of meta-divination is special onlyin that this aspect is clearly displayed at the surface of the ritual.

This is clearer perhaps in the Sisala case. A divination session could bedescribed as a ritual giving a certain picture of the ancestors and theirrelationship to the client. This picture could be judged on its own merit, asa more or less accurate description of the situation. However, the seconddivination suggests a very different idea: the justification for performing achecking procedure is that the ancestors can actually 'cheat' and distort amantic diagnosis; what the second ritual is supposed to establish is whetherthey allowed the first divination to be accurate. Thus, the combination of tworituals suggests that the diviner's diagnosis is not only a description of theancestors but also a consequence of their action (or abstention).

These remarks apply also to the other examples mentioned. In theMundang case, for instance, the series of ordinary questions, about theclient's situation, might be described as just giving a certain description. Theancestors are what the divination is about, and the diagnosis tells in whatstate the ancestors are as concerns the client and his or her endeavours. Thefirst four questions, however, suggest a different picture. Here the ancestorsare entities which cause the divinatory diagnosis. The fact that only they candisrupt it implies, conversely, that they are instrumental in making thediagnosis what it is in the end. This is also implied in the Tallensi case: theancestors are agencies which both bring about the diagnosis and aredescribed in that diagnosis. It is important to note that establishing thisdouble link, between the situation described and the diagnosis, does notnecessarily imply any intention on the ancestors' part. The Mambila examplegives a good illustration of that. Here the divinatory pots themselves are'asked' whether they are truthful; this is conceived as a mechanicaloperation. If the pot is truthful, then it answers the preliminary questions inthe right way, and that's that.

72


My claim here is that the value of mantic statements is evaluated on thebasis of an implicit causal criterion. Clients and diviners consider the ritualsas procedures which make it possible to obtain statements that are directlycaused by the situation at hand. A link is established between a certain stateof affairs (e.g., the ancestors' attitude to the client, or the fact that so and sois a sorcerer) and the statement which describes this situation. The implicitpremise is that the state of affairs itself makes it impossible for the statementto be otherwise; the hidden or unobservable situation causes the statementto be the way it is. Let us not forget that the link is established, in the client'smind, between an undefined and unobservable state of affairs, on the onehand, and a given assertion which describes it, on the other. My claim is thatthe clients trust the divinatory description because they assume that it iscaused by that undefined state of affairs.

To return to the French case studied by Favret and Contreras, if the queenof spades comes up rather than the knave of hearts, this event is linked to thefact that someone is busy trying to bewitch the client; and the link postulatedis, I think, of a causal nature. What is strongly implied in the divinatorysituation is that, given the human situation at hand, the cards could not havebeen drawn otherwise. Again, the participants imply that the situation,whatever it is, has provoked the specific patterns displayed by the cards.

The advantage of this interpretation is that it brings together the positiveaspects of both anthropological models of divination. In my discussion of the'intentional' paradigm, I pointed out that the exclusion of the actualspeaker's intentions is a crucial feature of divinatory communication. Nowthe problem is that the usual anthropological inference, that this exclusion issupposed to make room for another, hidden speaker's intentions, is plainlyfalse. We can see now why diviners are so explicit about the fact that theirintentions are immaterial. If my interpretation is correct, the divinatorytechnique is interpreted as a technique which allows situations to speak forthemselves, as it were. In this context, excluding the speaker's intentions canbe interpreted by the diviners and clients as a way of removing any obstacleson that path. This may also explain why, in many cases, the guarantees oftruth of divinatory rituals are mainly negative. What matters is that theancestors or the spirits or witchcraft will not disrupt the divinatory process.This can be easily understood in the causal framework presented here. Sincethe main idea is to let a situation cause an utterance about it, anyintervention, besides that of the technique itself, can only be conceived of asan interference.

Moreover, this interpretation allows us to understand the inescapablelimitation of the semiotic paradigm. It is certainly possible to describe omensor portents as ' signs' of certain situations, insofar as they are interpreted asstanding for them in certain circumstances. It should be added, however, thatthese signs must not be construed as signs in the ordinary sense, i.e., as

73


symbols, but rather as indices, to use Peirce's famous distinction. Theirrelationship to what they stand for is one of effect to cause.

This interpretation applies to both inspired and mechanical divination. Incases of inspired divination, where, for example, a priestess is supposed togive a message from the god, there is a direct causal link between the stateof the god, notably his alleged mental states, and the utterance about them.In mechanical divination, as we have seen, the situation itself is part of thecausal process which brings about the utterance concerning it. Both types ofdivination imply the idea of a causal link between the state of affairs peoplewant to know about and a description of that state of affairs. Moreover, thisinterpretation allows us to ignore another blurred anthropological dis-tinction, between divination as a technical operation and the interpretationof'natural' omens. Purely 'natural' omens, even if they are more of an idealtype than a proper category, seem to provide a simple example of the causalconnections described above. In the interpretation of omens, mostanthropologists accept a version of what I have called the semiotic model ofdivination. The main idea is that omens are interpreted as signs of things tocome, and that people who interpret omens are engaged in an operation ofdecoding. The idea of 'signs', however, is intrinsically ambiguous and doesnot make the link between signified and signifier precise enough foranalytical purposes. Again, what is sometimes described in anthropologicaltheories as a semiotic link, is certainly conceived of by the participants incausal terms. In the case of omens, the assumed causal relation links anoriginal state of affairs and two effects, only one of which is observable;omens are indirect evidence. The flight of the birds can allow one to predictthe demise of the empire because the state of affairs (whatever it is) that hasmade the birds fly this specific way is the one that will make the empire fall.8The idea of linking two effects to a single, undefined cause is in fact a verycommon procedure. When entering a room filled with smoke and with apeculiar smell, people would normally ascribe both to a single cause, even ifthey cannot identify what is the origin of both phenomena.

To recapitulate, I have tried to show that a precise description of the linkbetween situation and diagnosis may both solve the question of the veracityof divination and show the limitations of common anthropologicalinterpretations. In my view, when a divinatory diagnosis is asked about anunobservable situation, what is assumed is that, in certain circumstances, //is possible to make statements the content of which is governed by the verysituations they describe. This assumption, I would claim, is always implicit inthe participants' representation of the specific discourse situation ofdivinatory rituals. Besides this general assumption, subjects can haveextremely varied descriptions of the specific divinatory procedures availablein their cultural environment. In some traditional contexts there is a 'localtheory' (in fact some vague representations) about how divination works:

74


ancestors 'push' the diviner's wand, spirits force the beheaded chicken to runaway, and so on. These ideas constitute attempts to describe the causal chainwhich is implied by the very operation of divination. People may thusconceive a particular ritual as the reception of 'messages' from gods, spiritsor ancestors; they may also construe the divinatory technique as thedecipherment of'signs' and 'signatures'. In societies where personified godsare attributed all human characteristics, and supposed to be activelyconcerned with human affairs, it is no surprise that divination should beinterpreted as receiving messages from these anthropomorphic entities. Onthe other hand, when a group, like in the French Bocage example, hasvirtually eliminated the transcendent personnel, the causal principle is notexpressed in any other idiom than that of pure necessity: given a certainhidden situation, the divinatory diagnosis just could not be otherwise. Thesereasonings are sometimes made explicit, and accepted by most participants;they may also be mere individual rationalizations.

Such reasonings constitute the specific criteria of truth applied todivinatory situations. If the situations are described in those terms, then thestatements generated are likely to be conceived of as more reliable thanordinary discourse, which depends on the speakers' knowledge andintentions, among other factors. In spite of their obvious differences, thesespecific descriptions always presuppose the causal nature of the link betweena situation and its divinatory description. In other words, people assume thatthe situation, whatever it is, makes the divinatory statement be the way it is.This is the main point for a general anthropological account of divination.

Implicit causal concepts

The hypothesis presented here may seem complicated and 'expensive', inthat it attributes to the actors concerned a set of subtle hypotheses, aboutsituations, statements and their connections. This, however, is not the case.As we saw in chapter 3, people do not need explicit general hypotheses abouttruth and reality, in order to have specific criteria of truth. In much the sameway, people who take divinatory diagnoses as the consequence of thesituations they describe do not need theoretical elaborations. I am onlyclaiming that their representation of the divinatory utterance is, implicitly, acausal one, and in most circumstances it is bound to remain implicit orunconscious. This point is crucial to the explanation of divination as acriterion of truth, and will be extended to other forms of traditional discoursein the following chapters. It is therefore pertinent to determine exactly whatcognitive processes are implied here.

The anthropological study of causal thinking is somehow paradoxical. Onthe one hand, causal inferences are obviously central in the description ofcultures. Describing a specific cultural environment implies focusing, to a

75


large extent, on causal inferences that seem plausible in that environment andwould not be admitted elsewhere. It seems self-evident to the Fang that manycases of misfortune can be attributed to specific intentions on the part of thedeceased persons (bekong), a claim which would appear highly implausible toFrench Bocage diviners and their clients. A crucial task of any ethnographicdescription is therefore to explain how subjects are led to consider certainspecific causal inferences as plausible or natural. This question, however, isnot the subject of much anthropological reflection or speculation. AsNeedham points out (1976: 80), 'although the notion of cause has beencentral, since the nineteenth century, to anthropological disquisitions onprimitive thought, the modes of causality as conceived in alien traditionshave yet to be determined'. Although I cannot engage here in a detailedsurvey of anthropological ideas on the topic,9 it may be relevant to examinesome of the intellectual factors that led to this strange state of affairs.

Most anthropological ideas on causal thinking were elaborated in thecontext of the study of'primitive' or 'magical' thinking; they were thereforefocused exclusively on 'strange' causal inferences, e.g., the idea that recitinga certain spell can cure a disease or make the rain fall. In such contexts peoplemake causal connections which seem entirely different from those they findplausible in everyday life. It seems therefore plausible that there must beformal differences between these domains. The very concept of cause, and theprinciples following which an event A can be selected as the cause of eventB, must be different. The only difficulty with this approach is that it implies:

(i) that there is a strong, constraining concept of 'cause' used in everydaylife, with abstract principles which specify under what conditions eventscan be considered as causally related;

(ii) that we know what these principles are.The history of philosophical speculation on causality is a strident refutationof claim (ii); however refined the conceptions of causality, they neverencompass the variety of causal connections made in everyday life. As forclaim (i), cognitive studies tend to show that judgements of causalconnections are not made on the basis of abstract rules, about, e.g., thecontiguity of events or the fact that the connection considered can besubsumed under a general covering law. Rather, connections are hypo-thesised on the basis of their resemblance to core cases of simple, observablecausal processes, and of their compatibility with general empirical assump-tions. To take a very simple example, judging that the impact of the stonethrown at the window caused it to break implies resorting to empiricalgeneralities, e.g., about the relative breakability of glass and stone, besidesabstract considerations of spatial contiguity and chronological succession.On their own, such abstract principles would generate indefinitely manycausal connections which are never represented by subjects.10 All this tends

76


to show that anthropological discussions are wrong in assuming that'strange' causal inferences are rooted in a specific 'mode of thought',somehow divergent from everyday causal thinking. It is difficult to sayprecisely what such inferences diverge from. Everyday notions of causalityseem flexible enough to admit many 'strange' claims, provided there areenough background representations to make the resulting explanationsplausible.

Another defect of anthropological ideas on causality is that they focus onexplicit causal attributions, on situations for instance in which people statethat the rains came because the rainmaker performed the appropriate ritual.However, a study of the cognitive processes involved must take into accountthe fact that many everyday notions are implicitly causal. Understandingsuch basic notions as 'pushing', 'pulling' or 'killing' implies representingintrinsically causal relations. This point, as we will see presently, is directlyrelevant to my argument on divination and other traditional forms ofdiscourse.

To sum up, anthropological discussions of causality often neglect everydaycausal attributions and hypothesise specific cognitive mechanisms, specialisedin 'magical' and other such 'symbolic' reasonings. But the commonprocesses may well be powerful and flexible enough to accommodate most ofthe 'strange' connections people seem to find plausible. I will take thispossibility as a starting point in the discussion of traditional criteria of truth.Although what is said here is not meant as a thorough discussion of'magic',which is outside the scope of our problem, I hope it will convey the messagethat anthropological theories on such phenomena should be based on clearand plausible cognitive hypotheses rather than ad hoc constructions.

To return to our specific problem, I have suggested here that peopleinvolved in the interpretation of a divinatory statement evaluate it bypositing a causal link between the (undefined) situation at hand and thediagnosis that describes it. The process of evaluating certain descriptions orrepresentations, on the basis of implicit causal criteria, rather than descriptiveones, is in fact present in very simple everyday activities. A good illustration,which is often used in philosophical discussions, is the way we commonlyevaluate photographs.11 It may seem that a photograph of, e.g., KensingtonPalace counts as a photograph of that building because of its resemblance tothe original: the picture has the same parts, the same proportions and thesame ornaments as a given perspective of the real building. This would meanthat the picture is judged solely on its descriptive link to the objectrepresented, namely Kensington Palace. This, however, is not the case.Imagine the picture is taken with the wrong exposure, on the wrong type offilm, with the wrong lens, etc., so that it is difficult to see anything on theresulting image, let alone to see what the picture is a picture of. Such aphotograph would still be a photograph of Kensington Palace, although a

77


bad one. On the other hand, a picture of a model of Kensington Palace,whatever its resemblance to the real building, would never count as aphotograph of Kensington Palace. ' Being a photograph of is therefore anintrinsically causal notion, which applies only if the light reflected by theoriginal object has reached the film and caused chemical changes in it. Thisassumption is implicit in the ordinary understanding of the notion, even ifpeople do not necessarily know how light reaches the film or effects thosechemical changes.

In the interpretation of divination offered here, the participants'representation of divination is, in the same way, implicitly causal.Understanding the difference between ordinary and divinatory discourse, forthe subjects concerned, implies representing the latter as caused by thesituations represented. As we can see in the case of photographs, subjectswho assume a causal connection between an object or situation and itsrepresentation, do not necessarily have an elaborate 'theory' of theconnection. They assume that it exists, but do not necessarily represent whatit consists of.

The hypothesis of causal criteria has many implications for the study oftraditional truths. Let me mention one immediate consequence, the relevanceof which I will examine in the next chapter. In this account of divinatorytruths, truth-terms are applied, not to sentences or propositions, but tosingular utterances. I am referring here to the distinction between a sentence,i.e., an abstract reality generated by some grammar, and an utterance, aphysical realisation of the sentence.12 It is generally agreed that truthpredicates apply to sentences or rather to the propositions they express in agiven language, so that the idea of judging an utterance ' true' withoutapplying criteria of truth to the sentence expressed may seem strange. This,however, is a consequence of the special nature of causal criteria. When usingcausal criteria, people necessarily focus their reasonings on the utterance asan event. Worryingly, this seems to suggest that the content of the utterancesis not always the most important factor in the processes whereby they arejudged 'true'. If this is the case, then it should be possible to judge 'true' anutterance whose content is virtually eliminated. This possibility is indeedexploited in certain forms of traditional discourse, to which we shall nowturn.

78

Customised speech (II): truth withoutmeaning

Although traditions are commonly supposed to imply beliefs, conceptionsand world views, traditional contexts are difficult to describe in these terms.More often than not, the aspects of discourse which are linked to theexpression of underlying ideas and beliefs about the world are secondary orirrelevant, while the salient aspects are precisely not expressive in that sense.In chapter 2,1 examined a crucial feature of the categories used in traditionalinteraction, namely that they are anchored to memories of singular situationsrather than associated with mental definitions. As a consequence, expertutterances about these notions rely on the undefined resemblance betweensituations, rather than general propositions about the world. This is but anaspect of a general property of traditional discourse, which is ritualised in avariety of ways. Now, as Rappaport points out, an 'obvious aspect' of ritualand ritualised interaction is that they are non — discursive and non-explanatory (1914 passim).

I will now turn to a striking example of' non-expressivity' in traditions,namely the use of formalised or ritualised languages. In many traditionalcontexts truths are supposedly contained in utterances which the listeners,sometimes even the speakers, can barely understand. The discourse itself thusconstitutes an obstacle to the 'expression' of underlying 'ideas'. Suchphenomena cannot be just explained away as exceptions; in many societies,they are the hallmark of traditional truth. A theory of traditional discourseshould therefore account for this limiting case, where truth seems to becompatible with virtually meaningless utterances.

Ritual speech

Ritual speech may be construed as an intermediary form between two othermanipulations of language, viz. the construction of special languages and theuse of spells. Special languages range from argots and professional jargonsto the complex secret idioms used in some initiations and secret societies.They are used in various ways, to serve as identity markers or to make secretcommunication possible. They are generally formed on the basis of a natural

79


language, which is made incomprehensible to the outsiders by various formalprocesses, like morphological transformations, the systematic use of arepertoire of metaphors, a vocabulary either invented or taken from a foreigntongue, often in combination.1 Spells and other gnomic forms, on the otherhand, are an extreme form of the formulaic use of language. By virtue ofbeing repeated in a certain form or in certain contexts, some more or lessobscure formulae are supposed to have other properties, notably othereffects, than ordinary speech. In a sense, spells and secret languages aresymmetrical distortions of natural language. The point in using a speciallanguage is to create or emphasise differences in the code used, whatever themessage. Spells on the contrary stress the difference between differentpossible uses of the same language.

Ritual speech functions in both directions. On the one hand, it makes useof linguistic forms which are markedly distinct from ordinary speech; also,it is only used in specific contexts, where utterances are supposed to be moresalient than in ordinary conversation. As we will see in the examples andhypotheses presented below, ritual speech is usually described as moreefficient; its use in ritual contexts is often meant to have a material effect onsome situation. Most studies of ritual speech focus on this aspect. Here,however, I will deal with another puzzling aspect, with the relationshipestablished between ritual speech and truth. In most cases, what is said in theritual idiom, using the appropriate formulae, is supposed to convey sometruth which cannot be expressed in ordinary discourse. The problem then isto understand the combination of ritualisation and veracity.

A good illustration of this combination is Cuna formalised speech, asdescribed by Sherzer (1985). Among the Cuna of Panama, all knowledgebeyond common sense is embodied in various forms of ritualised discourse;knowing more than other people means knowing certain songs and formulae,and being able to master certain formalised styles. The Cuna emphasis onritual discourse is extreme; there is very little ritual activity besides theseformalised verbal genres. Cuna tradition as a whole is thus constituted bythree types of discourse and contexts, associated with the chiefs' oratory, thehealing rites and the girls' puberty rites respectively (Sherzer 1985: 22ff.).There are ritual specialists and a specific form of discourse associated witheach of these contexts; the special vocabulary used, as well as specificmorphological and syntactic rules and the introduction of parallelism, makeritual discourse virtually unintelligible for non-specialists; there are thusthree different special languages, which specialists learn by memorising songsand incantations.

The chiefs' discourse is especially interesting here, as it is supposed both toconvey important truths about the past and to set people good examples tofollow in everyday life. Chiefs only make this type of speech in the formalisedcontext of the common house (namakke) meetings; every other night the

80

Customised speech (II)

whole population of the village gathers there and listens to these speeches,usually structured as a dialogue between two chiefs. The subject matter ismainly mythological and cosmological, although it incorporates descriptionsof recent events or of the chiefs' experiences and various other stories. Thecontent of such dialogues is by no means stereotyped, and there is a gooddeal of improvisation; the style, however, is always that of formal oratory.The language used by the chiefs is barely intelligible to the listeners. After thechiefs' dialogue, which usually lasts for about two hours, two specialisedinterpreters give a shortened translation and some comments on the meaningof the narratives (Sherzer 1985: 72ff.).

In this context, much emphasis is laid on the truths contained in the chiefs'discourse. It must be stressed that this aspect is clearly conceived of asdistinct from the strictly political role played by the chiefs. In certain villagesthe functions are fulfilled by different chiefs; some of them, moreknowledgeable in mythological matters, make these ritualised speeches,while others are concerned with the management of day-to-day activities(Sherzer 1985: 57). Ritualised discourse is thus considered mainly as a sourceof true statements about the past and present states of village life. This aspectis of course difficult to combine with the fact that ritualised utterances aremade in a language which most listeners cannot understand at all. They areprovided with a translation, but the original is considered to be true, not thetranslation. Relevant aspects can be explained by the translators, but it isnecessary to resort to the formalised register beforehand.2

This has an important consequence, which in fact applies to most cases ofritualised speech: the utterances are supposed to convey a truth not in spiteof, but because of their formalisation. This of course goes against thecommon idea, that such forms of discourse consist in the expression ofworld-views or other cultural 'conceptions'. If we analyse ritual discourse inthis way, we are led to focus on aspects which are either nonexistent orirrelevant, while the aspects considered crucial by the actors, notably theformalisation, are just ignored. In a common conception of traditions, suchinstitutions are therefore incomprehensible.3 We must therefore examinewhether a different view would make it possible to account for this puzzlingaspect of traditional discourse.

The obvious problem with ritual speech is that it seems paradoxical. Themore formalised the utterances, the more they are conceived of as containingsome truth. At the same time however, more formalised utterances arenecessarily more difficult to understand, and this is often pushed to anextreme point, where the utterances are totally incomprehensible. How doesone get truth with less meaning or no meaning at all? If people's criteria oftruth are supposed to be descriptive, to be about the states of affairsdescribed by the sentences, there is no way to account for this paradox; onemust either ignore the facts or distort them. Here I will try to show that an

81


alternative hypothesis on criteria of truth could provide a solution to theparadox and account for the connection between truth and obscure ritualisedforms.

Formalisation and force: Bloch's hypothesis

The combination of formalisation and cognitive salience is the subject matterof an important argument put forward by M. Bloch (1974), which I shallexamine in some detail in this and the following section. Bloch's argument isfounded on an ethnographic analysis of two types of Merina ritual discourse,associated with political oratory and circumcision rites respectively. Theargument suggests that these types of discourse, although clearly different intheir aim and context, have many similarities; and that the mechanism ofpersuasion of religious discourse uses some of the 'coercive' devices oftraditional authority. I shall first present a brief summary of Bloch's ideas onritual languages, and then comment on some consequences of thesehypotheses.

Bloch's strong hypothesis is that the type of cognitive effect which isachieved by ritual statements is a consequence of formalisation. There is notjust a connection between these aspects, one is a consequence of the other.Formalised speech conveys less information about the world than ordinaryspeech, but it achieves another type of communicative effect, which dependson other aspects of the utterances, namely their 'illocutionary properties'.Formalisation can be seen as a modification of normal speech which reducesits information-conveying potential and by the same token enhances itsillocutionary potential; the effects of ritual speech on the audience areproportional to the reduction of the information transmitted. The idea ofcourse is directly opposed to what most theories of tradition would predict,on the assumption that traditional discourse expresses some underlying ideasabout the world, so that statements which express very little should not beconsidered by the actors as especially important and traditional. Contrary tothis, Bloch argues that both aspects, formalisation and salience, are not onlyfound together but are strongly connected. In his terms, 'with increasingformalisation, propositional force decreases and illocutionary forceincreases' (Bloch 1974: 67).

This idea of an inversed proportion is founded on a specific description ofthe formalisation of ritual languages. The main idea is that ritualformalisation or stylisation results in constraining the creative power oflanguage; not all sentences meet the prosodic or morphological requirementsof the ritual style. As a consequence, ritual languages are impoverishedlanguages. In Bloch's conception, the propositional force of a sentence isdescribed in terms very close to the idiom of information theory. Any givensentence is a consequence of some paradigmatic choices among sets of

82


possible sentences; the quantity of information conveyed by a sentence isinversely proportional to the probability of the choice made.4 The moreprobable (and consequently predictable) the sentence, the less informative.Stylisation is a reduction in the number of possible choices, so that theutterances are far more predictable than in ordinary speech; the stylisticconstraints are such that, from one sentence, the listeners can predict whatfollows; 'if this mode of communication is adopted there is hardly any choiceof what can be said' (Bloch 1974: 62). The loss of choice, and the consequentincrease of predictability, are an essential feature of any stylisation, andformalised ritual speech is an extreme example of this process.

In some cases the formalisation is such that there is no choice at all; givena certain point of departure, a speech-act A, and a certain compulsory style,speakers cannot but make the totally predictable speech act B. To use Bloch'simage, engaging in such a discourse is like being in a tunnel; having nopossibility to turn either left or right, the only thing one can do is to follow(Bloch 1974: 76). This allows Bloch to describe ritual language as intrinsicallycoercive. The actors are 'caught' in a discursive pattern which makes itimpossible to disagree or contradict, since the series of utterances ispredetermined from the outset. The choice for the actor, then, is betweenaccepting the conventions of discourse, thereby accepting the imposedutterances, and not speaking at all. As Bloch himself points out, this ofcourse is a limiting-case; formalisation is usually not complete; still, theextent to which it restricts the possibilities of a speech act B is ' the limitwithin which contradiction is possible' (Bloch 1974: 64). Ritual language canthus serve an ideological purpose, in that it is a 'hidden' yet powerfulmechanism which reduces drastically the possibility of dissent.5

Beyond the detailed analysis of Merina political discourse, Bloch'sinterpretation relies on some general assumptions that deserve a detaileddiscussion. Against most anthropological ideas on ritual discourse, itcontends that veracity, in such contexts, is a direct consequence offormalisation. The hypothesis seems far more plausible than commonanthropological ideas on discourse in ritual. The general premises, however,are not entirely satisfactory. In order to give a general and plausible accountof ritual speech, it is therefore necessary to examine the precise claims madeabout the cognitive effects of utterances.

First, let me mention some empirical problems concerning the applicationof the hypotheses. They concern a very specific type of situation, in whichvarious participants take part in what is presented as a debate but, due to theformalisation of discourse, does not in fact leave room for contradiction. Themodel should be qualified if it is meant to apply to cases where thedistribution of roles is entirely different, and where the tie-on of persuasionwith coercion is not so obvious. In the Cuna case, for example, the chiefs'ritualised discourse is supposed to convey some truth and it is highly

83


formalised. But the listeners, who take these utterances as true, are not meantto engage themselves in ritualised replies. In other words, there are cases(perhaps more representative of ritual speech than Merina political oratory)where the juxtaposition of ritual formalisation and cognitive effects does notrely on the fact that people are forced to 'participate' in the discourse, quitethe contrary. Shall we count out these phenomena as entirely different fromBloch's example? This would be odd, given the obvious and significantsimilarities. If we find in many areas of traditional discourse a combinationof formalisation and psychological salience or other such effects, this surelydeserves a general explanation. But then we must modify some features ofBloch's ingenious model.

The hypothesis of an inversed proportion is problematical, in that itsuggests that strong illocutionary effects are an automatic consequence of thereduction of propositional choice. This is obviously false, and many counter-examples can be found. In a conversation with someone who has a limitedcommand of the language, the choice of words is obviously reduced to themost common or basic expressions; utterances made in such conversationsare not especially 'strong' at the illocutionary level, quite the contrary.Indirect speech acts, irony and other such effects are ruled out. Therefore therelationship between these two aspects is more complex. Ritual formalisation,in Bloch's sense, must be construed as weak propositional choice combinedwith some awareness of other, non-reduced, types of discourse. Speakersmust be aware of what they could have said in a non-formalised style andcannot say in ritual language.

Even granted this common sense addition, the idea of a systematic relationseems problematical. It implies that a genre of ritual speech which is moreformalised is bound to have more effects than other types. But the evidencedoes not seem to support this. Very formalised utterances are sometimesfound with minimal effects; conversely, a highly inventive discourse canresult in strong effects. In the Fang syncretic bwiti religion, the mostpersuasive sermons turn out to be based on the invention of ingeniousmetaphors rather than on formalised talk (Fernandez 1986: 28-70, 172-86)and there are of course many such examples. This implies that Bloch'shypothesis is not false, but incomplete. The idea that formalisation entailstruth is insufficient, given the counter-examples, but it will give us the startingpoint of a plausible solution.

In order to go further, it may be of help to remember that the analysis ofdivinatory speech posed a similar problem. Divinatory statements are said tobe true because they are produced by the divinatory operations. Mostanthropological models just ignore this and try to describe the institution asthough statements were produced in spite o/the procedures.6 On the otherhand, the idea that clients find true whatever is said by diviners just begsthe question. What we must do is describe the additional assumptions which,

84


if admitted, make it possible to consider divinatory operations as truth-making. In other words, divination does not produce truth, unless one hassome implicit hypotheses about how it works. This, as we will see, is alsowhat happens in the case of ritual language. Veracity is taken by the listenersas necessarily linked to formalisation, and Bloch's model is an attempt totake this fact into account. But veracity would not be seen as a consequenceof formalisation if people did not have specific hypotheses about the way theutterances are produced. A satisfactory account of formal language mustdescribe these hypotheses.

Ulocutionary force: a problem of description

Bloch's account is founded on a distinction between propositional andillocutionary forces, two notions widely used in pragmatics. However, eachof these terms is taken here in a special sense, and I shall argue that someproblems in this explanation of ritual language stem from this particularconception of the interpretation of utterances.7 The force of ritual utterancescannot be explained in propositional terms; formalisation reduces the'propositional potential' of sentences to a minimum. In Bloch's model, theeffects on the audience, such as persuasion or coercion, must be attributed tothe illocutionary interpretation of the utterances. In this conception,illocutionary force is conceived as some kind of energy, something which anutterance can have more or less of, i.e., something which resembles 'forces'in the ordinary, physical sense. Now this is a far cry from most pragmaticaccounts of illocution.8 In pragmatic theory in general, illocution is not seenat all as a force in that sense; this point deserves some attention, as it iscrucial in the question of ritual speech. A fundamental idea of speech-acttheory is the distinction between locutionary and illocutionary aspects of anutterance. Sentences can be classified as declarative, interrogative, imperative,etc.; these are locutionary aspects. Now these aspects do not determine theillocutionary aspects, e.g., interpreting the utterance as a request, a statement,etc. Suppose one enters a very hot room, points at the window and tells thepeople sitting there: 'A bit stuffy in here, isn't it?' The sentence itself is acombination of a declarative and an interrogative sentence; these are itslocutionary aspects. Now, at the illocutionary level, the utterance obviouslycan be interpreted neither as a statement, as the other people are supposedto know that it is hot, nor as a genuine question. It is intended as an indirectrequest, something which could be also expressed by, e.g., 'would you mindopening a window?' and it is interpreted as such. The important point hereis that the illocutionary aspects of an utterance are qualities like, e.g., beinga request, an order, a wish, a statement. These are qualities an utterance hasor has not, they are not quantities it could have more or less of. An utterancecannot be more of a request than another one; it is or it is not a request. The

85


very term ' illocutionary/<?rce' is misleading, in that it suggests some variablequantity, whereas one is dealing with aspects which would be more properlydescribed as modalities.

Bloch's idea of an inversed proportion of propositional and illocutionary'forces' means that both can vary in quantity. This may apply topropositional potential, insofar as it is described in information-theoreticterms, but certainly not to illocutionary aspects. It may be of help here toresort to another distinction made by Austin, between illocutionary and'perlocutionary' aspects (1962: 100-7). In the example, mentioned above,making an indirect request is the illocutionary act; now the effects this act hason the people in the room are yet another matter. They might react, e.g., byopening a window, by protesting that it is not hot at all, or by makinganother indirect speech-act, saying for instance:' I happen to have a very badcold.' All such effects pertain to the perlocutionary aspects of the utterance.In the problem of ritual speech we obviously need some distinction of thissort. The utterances made have certain illocutionary aspects; they constituteorders, advice, statements and so on. What is of more interest to theanthropologist, however, is the effect of the utterances on the listeners. Ritualutterances are extremely salient, they are described as conveying some truth,they can be used to convince people. In his account, Bloch considers all suchaspects (together with ' illocution' in the usual, narrow sense) as part of the'illocutionary force' of an utterance.9 In order to account for the persuasioneffects of some utterances, Bloch needs to describe parallelism as a mechanismwhich precludes the possibility of alternative messages. But there are manycases of parallelism without this characteristic. So the persuasiveness ofritualised speech must depend, not only on the properties of the utterances,but also on the listeners' representations about the utterances and thedomain of reality which is talked about.

Utterances as events and instantiations

In both linguistic theory and everyday understanding, utterances can beconsidered from two quite different points of view. On the one hand,utterances are concrete instantiations of certain abstract objects, namelysentences, illocutionary forces, conversational maxims and so on. Anutterance, however, is also a simple singular event. Someone saying somethingat a certain location constitutes an event which is liable to be caused by otherevents or states of affairs. The Fang witch-doctor's assertion that ant-hillsmean witchcraft (made at a specific time before a specific audience) is aninstantiation of the sentence 'ant-hills mean witchcraft' which expressessome abstract proposition about a relationship between ant-hills andwitchcraft. However, it can also be described as a singular event that islocated in both time and space and can be memorised as such, as the person

86


so and so making the statement such and such in such and such singularsituation.

This distinction is relevant to the problem of parallelism and other suchprocedures used in ritual languages. Such transformations of naturallanguage have two consequences. They make it difficult to interpret theutterances as the expression of sentences, because of the lack of expressivepower described by Bloch. At the same time, some surface properties of theutterances become highly salient; parallelism displays at the surface of theutterances some syntactic or morphological or phonological structures whichare usually unnoticeable. The Kham-Magar shamanistic songs collected andstudied by A. de Sales in Nepal (1986) are a good illustration of thisphenomenon. They are sung by the shamans in a special language, with a verypoor vocabulary (not more than a thousand words, mainly of Nepali origin)and a simplified syntax, and make a constant use of parallelisticconstructions. Take for instance the following recurrent formula (Sales 1985;317):

Purai purkha na pura wastri ngaWhole husband you whole wife I

which is glossed as 'You are my legitimate husband, I am your legitimatewife.' In ordinary Kham-Magar language, this would be rendered by thesentence:

Na purai nga purkha nalizya, na purai na wastri ngalizyaYou whole mine husband you-are, I whole yours wife I-am.

In this example, the transformation from ordinary to ritual languageconcerns both morphology and syntax. There is no verb in the ritual formula,and the juxtaposition of the elements in a single line is supposed to replacethe copula. In the same way,

Naza sahina na naza bahina naYou you-can you, you you-speak youis glossed as 'you are the one who can speak', a proposition whosecomplexity is far beyond the possibilities of the ritual syntax. As Sales putsit, words in the ritual language seem to be 'handled like objects' (Sales 1985:317), as independent units that can be juxtaposed or separated for reasonsother than syntactical.

These are very common characteristics of ritualised speech. In the space ofthis chapter it is not possible to present a detailed description of parallelisticand other distortions, or to enter into the description of their cognitiveeffects. Two points, however, must be noted, because they are directlyrelevant to the question of truth. First, listeners who do not master the basicvocabulary of such songs recognise them as series of utterances rather thannoise only by virtue of the strong prosodic organisation; this poetry is always

87


'opaque' in the sense that it is impossible to interpret it without focusing onsurface properties like, e.g., the recurrence of certain phonologicalcombinations.10 Second, inasmuch as listeners have some knowledge of thelimited vocabulary, they are bound to focus on the iconic aspects ofparallelistic constructions. There is a strong, inevitable element of iconicityin ritual speech, if only because the vocabulary and syntax are not richenough to express all the thoughts conveyed. Thus, Sales points out that theunity of the mythic couple, in the sentence 'Whole husband you whole wifeI,' is suggested by the syntactic structure itself, by the juxtaposition of thewords (Sales 1985: 318). Some features of the objects described aresupposedly reflected in the surface properties of the utterances. Predicationis often conveyed by contiguity, opposition by prosodic breaks, and so on.As a result, the listener's interpretation of ritual speech, if the latter is treatedas something other than random noise, is necessarily focused on theproperties of the utterances as phonic events, as material occurrences. At thesame time, the obscurity and ambiguity of formalised utterances make itmore difficult to interpret them as the realisation of abstract objects, notablyof sentences. If we take this into account, it may be possible to completeBloch's model and put forward a general account of truth in formalisedlanguage.

Ritual speech as caused utterances

In chapter 4,1 put forward the hypothesis that divinatory operations can betaken as an argument for truth only if the listeners assume that the diagnosisis caused by the very situation it describes. In other words, the fact thatdivination is persuasive is not a direct consequence of its formal properties;it is a consequence of the availability, in the listeners' minds, of a causaldescription of these formal properties. People do not judge divinationtruthful simply because it is divination, but because they have an implicitcausal assumption, in which the statement is caused by the undefinedsituation at hand.

Applying this interpretation to the case of ritual speech would imply thatthe persuasive effects are not attributed to parallelism as such, but to the factthat people interpret formalised utterances as caused by the very states ofaffairs or entities they describe. Formalisation would be represented bypeople as an index of the fact that the utterances produced are influenced ortriggered by ancestors, witchcraft, spirits, etc. This assumption can beentertained only if the listeners have representations about both the existenceof a causal link and the fact that formalisation is its index. They must have(i) some representation of the fact that the ancestors, spirits and other entitiesdescribed are actually instrumental in bringing about the utterancesconcerned, and (ii) some representation of the surface properties of theformalised utterances as a consequence of this process.

88


Bloch's own description of Merina ritual oratory seems to bear out thiscausal interpretation. The rituals begin with some elders explaining thepurpose of the ceremony, in a style which is closely similar to that of ordinarypolitical oratory. A notable change occurs, however, as the ritual proceeds.The speeches delivered after a few hours are very different from politicaloratory, in content as well as style; they tend to be indistinct and moreparallelistic than at the beginning. 'They would be classed [by the observer]as the speeches of persons under possession. The elders appear to be in atrance-like state...' (Bloch 1985: 58). At no stage, in these speeches, are theelders supposed to speak on their own behalf; what they are saying are ' thewords of the ancestors'. People apply this description to both the speechesof the beginning, which sound like political oratory, and the subsequent,'possession-sounding' discourse. At both stages, the words of the ancestorsare supposedly realised by the elders' utterances: ' for them possession is anextreme form of saying what would have been said by the elders of the past,or repeating their words' (Bloch, 1985: 59). There is a marked difference,however, between the two stages, as Bloch points out: 'instead of theancestors speaking indirectly through the memory of the living elders theyspeak directly through their person' (1985, Bloch's emphasis).

This type of change seems to suggest that the elders' utterances graduallybecome more directly influenced by the ancestors themselves. Instead ofrepeating what they think were ancient sentences, the speakers are forced toutter what the ancestors make them say. Bloch emphasises the fact that theconnection between the ancestors and the elders' speech gradually becomesa direct one. At the beginning of the ritual, the elders are supposed to bequoting what they can remember. As the ritual proceeds, however, they areby-passed, as it were, in a process which links what the ancestors want to sayand what the elders actually utter. This is an obvious feature of possession;but I must insist on the fact that in such a form of discourse, what is said issupposed to be directly caused by some entity other than the speakersthemselves. If this interpretation is correct, the formalised utterances aresystematically interpreted as directly influenced by events or states of affairswhich they describe. Most formalised discourse should then be construed asnot only a quotation but, more strongly, as the consequence of the realitydescribed. Instead of being a quotation of, e.g., 'the words of the ancestors',formalised utterances in ritual contexts would be interpreted as triggered bythe ancestors.

It is important at this point not to misunderstand the ' quoted' aspects ofritual discourse. In the Merina case, for instance, it seems that the listenersnaturally interpret the formalised utterances as simply spoken by ' someoneelse', namely the ancestors, through the elders' mouths. Then formalisedspeech would be just quoted speech, the irruption of some alien speaker insomeone's discourse. This is suggested by Sherzer in his interpretation ofCuna ritual speech, which systematically includes reflexive and meta-

89


communicative elements, so that utterances are ' one or more steps (actuallyone or more mouths) removed from the actual speaker' (1985: 210). I do notwant to query the relevance of this interpretation in the specific casesconsidered; taken as a general principle, however, the hypothesis would befalse.

A parallel with divination will make this clearer. As we saw in chapter 4,the clients in certain forms of divination assume that the diagnosis is notproduced by the diviner but have no idea who then is 'speaking', and oftendo not even bother to consider this problem. What is assumed is that, giventhe (undefined) situation at hand, the diagnosis could not have beenotherwise. As I said, the only assumption here is that the diagnosis is causedby the situation. The idea of hidden or divine 'speakers' is not necessary tothis causal interpretation; in those cases, however, where the subject-matterof divination is the influence of personified agencies like gods and spirits, itis quite natural to interpret the diagnosis as the words of these gods andspirits. The diagnosis is then interpreted as caused by these agencies'intentions. This, however, is just one possible way of describing a causal linkbetween situations and utterances; there are many forms of divination inwhich a causal link is assumed, without this 'personified' description.

In much the same way, we must be wary of an interpretation whichdescribes the ritual speaker as just a 'spokesman' for spirits or ancestors.There are cases in which formalised speech is supposed to contain sometruths, without any hypothesis about the ' hidden speakers' behind the actualones. The idea that ritual speech is interpreted as the words of mysticalentities is therefore too specific, if we want to account for this phenomenonin general. Both personified and non-personified interpretations of ritualspeech presuppose the idea of a causal link from an undefined situation to theutterances actually made.

To sum up: what is proposed here is that formalised or parallelistic speech,in these ritual contexts, is taken by the listeners as a form of speech directlyinfluenced by some situation or some entities described. Pace Bloch, I wouldclaim that 'reduced propositional potential' on its own, is not sufficient forthe utterances to be held true. Semantic vacuity and parallelism are indicesof truth only if the listeners have some assumptions about the reasons whydiscourse is thus distorted. Now parallelism and other types of formalisationfocus the listeners' attention on the properties of utterances as events, andtherefore make it possible to entertain hypotheses about the utterances asevents caused. This, of course, is only one of the effects of parallelism, but itis necessary to take it into account if we want to understand why and howsuch discourse can be held truthful in spite of its reduced 'propositionalpotential'.

90


Truths as a rare commodity

Divination techniques and the ritualisation of speech obviously pose differentproblems; in both cases, however, traditional truths are created by'customising' speech. The production of ordinary discourse involves differentmechanisms, and the customisation consists in substituting artificialmechanisms for the natural processes at some point in the production ofutterances. In divination, this applies to the assumption, following which thecontent of a statement is usually determined by what a person wishes toexpress; in such contexts, the speaker's intention is replaced with a technicaloperation whose result is unpredictable. In the case of ritual speech, thisapplies to another usual assumption, namely that statements can beunderstood as describing a certain state of affairs; this principle is demotedand the utterances are consequently represented as patterns of surfacefeatures. There are no doubt many other ways of customising discourse;these examples are only two illustrations of the possibilities and results of thisprocess.

The idea of truth criteria being applied to utterances as events caused,rather than sentences expressed, makes it possible to understand animportant feature of the traditional style, namely its 'literalism'. In chapter1, I used the term in an intuitive way, to designate the fact that people wantstatements made in traditional contexts to be repeated accurately. Thisemphasis on accuracy implies that a paraphrase of an utterance cannot havequite the same value as a verbatim repetition, though both have the samecontent. What I called 'expert utterances', in Fang discourse, are generallycited in their exact phrasing. People feel that any change in the surfacefeatures of the utterance would affect its value. Indeed, the very idea of aparaphrase, of a change that does not affect the truth conditions of thesentence, seems alien to the traditional style. A classical interpretation ofthese aspects of traditional style is founded on the idea of traditionalism; thismay be described as a preference for time-tested conceptions or as aninstinctive compulsion of repetition. In chapter 1, I mentioned some of theproblems created by this type of psychological explanation, in which peopleare said to stick to old formulations because they want certain ideas orconceptions to remain unchanged. Applied to the specific question ofliteralism, such a conception would imply that people prefer verbatimrepetition because it is a safe way of preserving the ideas or conceptionsexpressed. But in the limiting case of ritual speech, people are 'literalist'about utterances which have very little 'propositional content', to useBloch's terms. These are utterances about which it is certainly impossible torepresent what the 'content' to be conserved is, since there is virtually none.So literalism cannot be construed as the desire to conserve 'conceptions' and'meanings'. In the model presented here, there is no need to suppose that

91


people represent some 'conception' which they wish to preserve. Literalismcan be understood as a consequence of the choice of causal criteria of truthin certain contexts. Utterances are then considered true insofar as the eventof the utterance meets certain requirements, notably if it is possible torepresent a causal link between the situation described and the utterance. Ifsuch criteria are taken into account, the only reliable utterances are those forwhich a relevant causal description is available.

An important consequence is that, to use a convenient metaphor, truth isa rare commodity. The number of situations in which it is necessary to finda true description or explanation of a state of affairs greatly outnumbers thenumber of contexts in which guaranteed truths are produced. There are onlya limited number of divination operations or ritual speech sessions, and mostof the truths produced only concern very specific questions. So the problemis to extend the applicability of ritually guaranteed statements to othersituations. Ideally, this could be done in two different ways, either byinferring new truths on the basis of the ritually guaranteed ones, or byseeking some analogies, between the object of the ritual truths and the newsituations at hand. The problem with the first strategy is that it does not offerany guarantee about the truth of what is so deduced. As I pointed out, causalcriteria can only apply to singular utterances. Once inferences and deductionsare made, there is no guarantee that they will be as true as the originalstatement, because there is no causal link between them and the situationdescribed. The second strategy, seeking resemblance between situations,while quoting the same singular truths, is the very principle of literalism. Letme return again to the example of Fang ant-hills and witchcraft mentionedin chapter 2. The statement was made in such circumstances that it wassupposed to be true. Now there are many situations where certainties aboutwitchcraft are in high demand. In such situations, people do not try toextrapolate some general principles from the specific utterance about ant-hills ; what they do is quote this original statement, although the situation athand is quite different. The implicit idea is that there must be someresemblance between the two situations, although it is not precisely described.

As I mentioned above, an utterance can be apprehended and described intwo different ways, as an instantiation of some abstract object (a sentence)or as a singular speech-event. Both aspects are represented in the listeners'interpretation; an utterance is construed as realising a certain sentence, andit is also an event which has causes, effects and circumstances. As aconsequence, literalism is manifest in two typical ways. Quoting is the mostobvious one; people tend to favor the repetition of an utterance rather thanadapting the original to the new circumstances. Quoting an utterance ismaking another utterance which shares the instantiation properties of thefirst one; it is a realisation of the same sentence, and consequently people canimagine that it has the same truth-value as the original one. I would argue,

92


however, that due to the existence of criteria focused on events, the quotationof an utterance cannot be considered as obviously true as the utterance itself.This is why another common property of traditional contexts is thereiteration of the circumstances which guarantee the truth of the statements.Reiteration focuses on the event properties of the utterance; if a truth hasbeen produced by resorting to a certain technical operation or to a certaintype of specialist, then an obvious way to create new truths is to resort to thesame type of operations and specialists. In this way, one reproduces theproperties of the utterances which make it possible to imagine a causal linkbetween the states of affairs described and the speech-events.

Literalism is, then, more complex than is usually assumed in an-thropological descriptions. It does not involve any specially 'conservative'tendency, but is a consequence of the type of criteria used to judge thatstatements are true. It is also more complex in the sense that two differentphenomena are concerned, the prevalence of reported speech and thetendency to reiterate certain circumstances for the expression of truths. Thesetwo aspects are not usually described together, although they seem to be twoconsequences of the same cognitive processes involved in the interpretationof traditional statements.11

93

Customised persons: initiation,competence and position

A salient aspect of traditional situations is that the access to truth, moreprecisely the capacity to make true statements, is reserved for persons placedin certain social positions. If one considers traditional statements as theexpression of transmitted 'theories' or 'world-views', there are two ways ofexplaining why truth-bearing discourse is thus reserved for certain actors. Thefirst solution is to have an 'authoritarian' idea of the groups concerned, i.e.,to claim that in such groups whatever is said by the 'powerful' is held true.The second possible solution is that the actors concerned have a specialcompetence, notably a greater knowledge of traditional 'theories' thanordinary speakers. At the end of chapter 3,1 mentioned the problems createdby the 'authoritarian' interpretation. The main difficulty is that the claim justbegs the question of the intellectual processes whereby a person is endowedwith traditional 'authority'. If position is a simple criterion of truth, why andhow is it considered a criterion? If, for example, shamans are the only oneswho can say the truth about the spirits, what are the listeners' representationsabout this capacity? Apparently, the only reasonable answer to this questionis some notion of 'competence'; the privileged speakers have moreknowledge than the others. This is the most natural hypothesis; in fact, if theveracity of utterances is evaluated on the basis of their content, it is the onlyreasonable hypothesis. As we saw in the previous chapters, however, thecriteria used in the evaluation of traditional statements are not that simple.They are based on representations which, to some extent, 'by-pass'considerations of content. Indeed, there are limiting cases where content isvirtually absent, as we saw in the examination of ritual languages.

Clearly, the idea of'competence' is more problematic than it first seemed.This is why I will now turn to institutions which are supposed to give somepeople the capacity of stating the truth, yet are not amenable to a descriptionin terms of 'competence'. This in turn will help us make some hypothesesabout the way social positions are represented in traditional interaction.Traditional truths often require a 'customisation' of speech; they alsorequire the special preparation of persons. Here I will consider the mostsalient example of such preparations, namely the rituals which supposedly

94

Customised persons

create 'new' persons and consequently make them capable of truthfuldiscourse.

The cognitive paradox of initiation and competence

Initiation rites are not easily defined, but their ' family resemblance' is basedon a combination of (i) a change of status, supposedly effected by the ritualitself and often expressed in terms of knowledge or access to someinformation, and (ii) some formal characteristics such as the seclusion ofcandidates, the secrecy surrounding initiation proceedings, the painful orhumiliating ordeals, the use of specialised languages or codes or other meansof representations, etc.1 One of the main anthropological problems is that ofthe status of initiation knowledge. In many societies initiations are said toconfer some kind of knowledge, either practical (hunting, fishing, etc.) ormystical (some secrets about ghosts or ancestors or masks). In either case itis difficult to take such presentations of initiation at face value. Mostactivities 'learned' were in fact largely known, or even practised, by thecandidates before their admission. And whatever mystical knowledge istransmitted seems to focus on secrets either meaningless or empty. Forinstance, neophytes are told that the masks, previously described asmysterious instantiations of ancestors or spirits, are just a gross deception;they are actually worn by fathers, uncles and other male adults.2 In somecases the communication of secrets can be delayed indefinitely. In theBaktaman initiations described by F. Barth (1975), candidates proceedthrough a system of grades; what they learn at each step is that the essentialsecrets, the ultimate explanations, will be given next time. In many initiationscandidates have to learn a special language or system of signs, supposed toembody a new, deeper understanding of the world; most of these newsemiotic systems boil down to simple rhetorical modifications of the nativetongue and the common representations, so that what is learned is eithertrivial or empty (Boyer 1980: passim).3 Not to put too fine a point on it,although initiations do have important cognitive effects, it does not seem thatinitiates ever know anything more than the others (Richards 1956: 126).

Initiation rites thus display what may be called a cognitive paradox. Acommon feature of initiation operations is their obscurity; they are difficultto describe as anything but gratuitous vexations and the solemn impositionof a vacuous 'knowledge'. It is quite common that initiates have nothingmuch to say about what they had to go through, not just because it isforbidden to communicate those secrets, but also because their memories ofthe rite do not constitute any significant description. Thus, a change whichis most often expressed in cognitive terms is brought about by procedureswhich seem either impossible to describe or too trivial to mention. What isdifficult is to bring together both aspects of initiation rites, the supposed

95


effects and the procedures. It may seem possible to bridge the gap by refiningour idea of the kind of knowledge acquired in such contexts; this is the mainpoint put forward by J. La Fontaine for instance (1977: passim; 1985:141-61), following whom it is an oversimplification to reduce initiation'knowledge' to information transmitted. In the 'tribal' rituals of the Gisu ofWest Africa, what seem to be transmitted are skills of various kinds ratherthan knowledge properly speaking. The difference between initiates andneophytes is commonly expressed in terms of behaviour; the latter havesupposedly acquired a certain ' know-how' as regards the relationships withancestors or the management of social intercourse. Obviously, such acompetence could not be easily reduced to explicit 'lessons' and thetransmission of information; they rather imply an apprenticeship based onexample and imitation. Initiation does not bring new 'ideas' but the masteryof new procedures.

However pertinent in the ethnographic context considered, these ideascannot really solve the paradox mentioned above, for they imply a somewhatdistorted view of those two aspects whose combination is problematic. Thatinitiations can confer some kind of 'know how' is certainly true; however,it remains that in the groups concerned, the difference is expressed in termsof secret knowledge and implies that the initiates' utterances on certaindomains are true. And the problem remains, as concerns the procedures;even if we admitted that initiations transmitted some 'competence', the typeof procedures used would remain quite puzzling. Competence can surely betransmitted through example and imitation, but initiation ordeals areprecisely difficult to describe as meaningful 'examples' of anything, and veryoften there is no relationship at all between what is experienced during theritual and the type of competent behaviour expected afterwards.

Let me illustrate these points with an extreme case, that of male initiationamong the Beti of Cameroon; the rite is mentioned by many ethnographersin the area because of its outstanding complexity and symbolic elaboration(see Houseman 1984, 1986). It involves several months of secluded life inbush camps, with a complicated series of ordeals, the main feature of whichis a combination of logical quandaries and humiliation or physical violence.Most of the ordeals could be described as the enactment of a conceptualantinomy. Thus, the candidates are told to go hunting some invisible animalsin the bush, and find themselves hunted and attacked by the elders. They aretold about the delicious fat of the antelope (soo), a delicacy they will be givenat a crucial point in the ritual; and in fact they are force-fed a disgustingmixture of rotten food, excrement and semen. They are told they will learnhow to forge iron, and consequently must 'practice', which in fact meanshaving their fingers crushed between heavy logs. In much the same way,'blowing the initiation horn' actually means being forced to blow in themaster's anus, ' being allowed to wash' means being forced to roll on thorny

96

Customised persons

bushes, and so on (Houseman 1984, 1986). Most Beti men have gone throughthese ordeals, and anyone who has failed them would find it impossible tohold any office or have any position other than that of a social minor. Thelack of a parallel institution is the official reason of women's alleged inabilityto take any important decisions or indeed make any definite and importantstatement about things that matter, such as marriage, witchcraft, or villagepolitics.

The Beti example provides a dramatic illustration of the cognitive paradoxof initiations. While the ordeals are incomprehensible (and unpleasantly so)the resulting 'competence' is maximal, and confers the capacity of saying thetruth on most domains of reality. In some traditional initiations the contrastis not so evident, but the problem remains, that initiations are defined asrituals which, among other things, allow some people to make truestatements about certain matters. The actors concerned seldom explain howgoing through a humiliating ordeal should make one more 'competent'; thisis why anthropological studies tend to play down the problem, or reduce itto the construction of an impressive decorum. But this implies a massivedistortion of the ethnographical facts. All the groups concerned emphasisethat the ordeals are a condition of competence sine qua non. As La Fontainepoints out, this is taken as an axiom. Our task then should be to explain howthis axiom can be taken as convincing or inescapable.

Initiations and causal interpretation

In the common implicit view of traditional truth, members of the groupsconcerned are supposed to think that the initiates have a 4 correct' picture ofthe world, which is communicated in their utterances; and this is why theutterances are taken to be true. As this view of initiation truth leads to theparadox mentioned above, we must examine whether a causal account wouldfare better. In such a framework, the initiates' utterance would be consideredtrue by the listeners insofar as they can represent some causal link betweencertain domains of reality and the utterances made about them. Initiationrituals would then be situations whose representation allows people torepresent such a causal link.

This account seems to follow very closely the very interpretation the actorsusually give of their initiation rituals. Let us consider for instance the case ofthe Mianmin initiations described by D. S. Gardner (1983). The successivestages of rituals establish the formal distinction between dill and duwaiin,those who know and those who don't. The rituals include the presentationof ancestral relics (skulls and trophies) as well as the revelation of the secretnames of plants, animals and artefacts. Neophytes are also taught secretsongs, generally of a mythical character. In short, everything here seems tobe expressed in terms of information communicated. However, Gardner

97


points out that such a description must be qualified, as ' rituals are consideredefficient regardless of what can be called the cognitive change brought aboutin the candidates' (1983: 352). The rituals are successfully accomplished,even if the candidates have not understood anything in the initiatic teaching.When the relics are exposed, it is not compulsory to look at them. As for thesecret names of things, one is allowed not to hear or memorise them; thisdoes not affect the general result of the ritual. Not to put too fine a point onit, the crucial element in the ritual is not the transmission of informationabout the ancestors, but the fact that the candidates have been ' exposed' tothe power or influence of the ancestors. This 'exposure' is what makes thedifference between diil and duwaiin (1983: 339); this is why the initiaticprocess is viewed as a causal one, which brings about changes in people byfollowing some recipe, even if people have no idea how the process actuallyworks (1983: 353).

This I think is not an exceptional example; in most initiations, the fact thatinformation is not properly communicated does not in any sense hinder theritual process. Indeed, in many cases the candidates actually avoid receivingany information at all, e.g., in the Bambara /romo-brotherhood rituals inMali, in which most neophytes stop their ears when being told the potentiallyharmful 'secrets' of the brotherhood (Dieterlen and Cisse 1972: 80). In theBaktaman rituals described by F. Barth, the candidates develop an'epistemology of secrecy', the main principle of which is that ritualknowledge is dangerous, so that it is safer to avoid it as much as possible(Barth 1975: 218ff.). What is always assumed and proclaimed in suchcontexts is that there is a necessary link between the ritual process itself andthe veracity of subsequent assertions; being initiated is a prerequisite oftruthful discourse. This idea of a necessary condition is puzzling if we assumethat people evaluate the initiates' assertions in a 'descriptive' way. Ifutterances are judged solely on their content, then whether they wereacquired in a specific context or not should not change their value. The factthat they were acquired in paradoxical, unintelligible ordeals should not addto their reliability, quite the contrary. As we saw in other domains, however,criteria of truth are not necessarily founded on such a 'descriptive'evaluation. Like any form of representation, utterances can be judged ontheir causal links to the situations they describe, i.e., as consequences orindices of these situations. This seems to be the case for the criteria appliedto many traditional statements. Instead of being judged only on theircontent, utterances are also considered as events which, like any other typeof events, have causes and consequences. This makes it possible to considerthe statement about a certain state of affairs as a consequence of that stateof affairs.

I would argue that it is possible, in this 'causal' framework, to solve thecognitive paradox of initiations. In this view, what the initiates assert about,

98

Customised persons

say, ancestors would be represented as something caused by the ancestorsthemselves. The features of initiations we considered here precisely suggestthat one of the aims is to make it possible for hidden entities and agencies tohave some direct effects on the neophytes. I would therefore argue that, ifinitiation rites are conceived by the actors as consisting in such contacts anddirect encounters, they make hypotheses about subsequent discourse astriggered by these agencies much more salient. Initiation rites, among otherthings, make it possible to represent the initiates as persons under directinfluence, therefore as persons whose utterances are more likely to be causedby the agencies, entities or states of affairs they make statements about.

In the light of this hypothesis, the main features of initiation rites becomemuch less puzzling and paradoxical. The fact that the information conveyedduring the rites is virtually vacuous, and that the candidates can avoid theteaching and still be considered initiates, all this is difficult to understand ina 'descriptive' account. If we admit that people use causal criteria, on theother hand, the transmission of information clearly becomes secondary.What makes initiates more 'truthful' than ordinary speakers is a series ofevents, in which they are put in direct contact with hidden agencies orentities. Clearly, if this is the point of the rite, the acquisition of informationis irrelevant, or else is considered as a mere side-effect of the direct contactestablished. Obviously, this is not all there is to say about initiation rites andtheir recurrent features, notably paradoxical or impossible ordeals, the use ofempty secrets and physical violence.4 Our problem here, however, is not toaccount for these striking features, but to understand how they can beconceived as a prerequisite for truthful discourse. A general account ofinitiation rites and their cognitive effects would certainly require more refineddescriptions and hypotheses; but I would claim that, inasmuch as it dealswith initiation as a condition of truth, it would have to take into account thefact that the initiates' utterances are judged on the basis of causal criteria.

Competence and position

A consequence of our causal account is to change drastically the idea of'competence' or 'knowledge' involved in such statements. It is alwaystempting to view traditional specialists as the equivalent of experts incomplex societies, i.e., people especially competent in some domain.However, the ordinary idea of competence implies that some people haveacquired an accurate representation of a certain domain of reality, so thattheir utterances describe what the domain really is. The idea that someone iscompetent about Italy because they have travelled there extensively rests onthe assumption that travelling there has given them an accurate rep-resentation of the country. Contrast this with the case of utterances judgedon the basis of causal criteria. What is established as a link between the

99


realities described and the utterance; as a consequence, the representationswhich make the idea of this link more salient are bound to be used as criteriaof truth. The 'customisation' of speech provides such representations, and sodoes the customisation of persons which initiation rites purport toaccomplish. Divination in itself is not a guarantee of veracity, no more thaninitiation rites in themselves make people truthful. The point is that suchelements can become criteria of truth only if the listeners have specificassumptions about the way the utterances are produced, e.g., the idea thatutterances are directly caused by the situations at hand. In such cases, thefact that the experts have an accurate representation of the world is by-passed, as I said above. Situations cause utterances by speaking forthemselves, as it were. If this is the way the listeners interpret a series ofutterances, then the representations stored in the experts' minds are notsupposed to play any role in the production of utterances. This, in fact, iswhat many speakers in traditional contexts forcefully argue, by claiming thattheir utterances are 'inherited' or 'inspired', i.e., not a consequence of theirown knowledge or communicative intentions.

The idea of traditional 'competence', as found in anthropologicaldescriptions, is inadequate to describe such processes; its main flaw is that itis ambiguous. If the claim is taken in a strong way, to imply that traditionalexperts derive the utterances they make from general knowledge of a'theoretical' nature about the topics treated, then it is clearly false; the ritualswhich are supposed to convey this knowledge turn out to be vacuous, thesecrets are empty. If, on the other hand, the idea of 'competence' onlyimplies that experts know the precise words of the myth or the gestures of theritual, then it is certainly true but is not sufficient in order to explain why theutterances should be considered true. In the causal account I have putforward, there is no need to suppose that the listeners interpret the veracityof the utterances as an effect of the speaker's knowledge. An utterance isjudged true, not because the speaker has the appropriate picture of the worldin his or her mind, but because a causal link is assumed to exist between thestate of affairs described and the event of the person making the utterance.

If we accept this interpretation, then we must have some precise hypothesesabout the way social positions are represented in traditional interaction. Weare claiming that people represent a link between a state of affairs and theutterance about it. But if only certain persons are supposed to make truestatements, then it should imply that only these persons are considered liableto be involved in such causal processes. To take a simple example, theaudience consider that shamans speak the truth about spirits because thespirit-situation, as it were, speaks through them; but the audience also thinksthat only shamans say the truth about spirits. It follows that there must besome representations which make it possible to think that only shamans canbe involved in the causal process in question. These persons are different;

100

Customised persons

rather, people have some representations about their being different, whichaccount for the fact that only they can have situations speak through them.What are these representations?

Traditional positions, i.e., positions as implied in traditional interaction,are usually taken as a subset in the larger domain of social statuses, definedby specific combinations of rights and duties.5 The problem with thisapproach is that it deals with an objective evaluation; it is very unlikely thatsocial positions are represented in that way. Among the Fang, positions suchas 'ngengang' (healer) or 'mbommvet' (epic singer) are more than'complexes of roles' objectively defined in terms of rights and duties, becausethey are named; it is the underlying mental processes of categorisation thatwe must explain here.

Terms used to denote social positions are usually characterised by the factthat the persons to whom they apply perform certain activities. If there wasa systematic link between category membership and activities, it would bepossible to consider traditional positions as some kind of system. However,the link is not systematic at all. There is not only a problem about someactivities being shared between several positions and vice versa (this wouldnot necessarily compromise the idea of a system or of structured semanticdomains); a more difficult problem concerns the very nature of the link. Theidea of a definition implies that the activities constitute a list of conditions formembership in the category concerned.6 In the following sections I willexamine whether the occupations used to characterise positions are necessaryor sufficient, and then draw some conclusions about the way positions andpersons are represented in traditional contexts.

Positions and their * definition9

Let me first consider the question whether the occupation cited as thecharacteristic of a class of people is a necessary feature of that class. Let ustake the example of the 'castes' of craftsmen that can be found in most WestAfrican societies. The usual pattern is that of a small group of people whoundertake all non-agricultural activities (blacksmithing, pottery, weaving,medicine, etc.) opposed to a vast majority of agriculturalists. The craftsmenare generally deemed 'inferior', 'polluting' and isolated from the rest of thegroup by a strict endogamy (Clement 1948, Maret 1980).7 There is a complexsymbolic elaboration about the exclusion and impurity of these people, andtheir activities are central in these reasonings. The Mafa of Cameroon forinstance have a caste called ngwalda; these are the group's blacksmiths,potters, healers, undertakers and diviners. The exclusion of these people isjustified by the fact that they manipulate such dangerous things as iron-oreand fire, that they handle polluting objects like corpses, that they bake piecesof earth (pots) rather than planting things in the earth to 'be baked' (i.e., to

101


grow) there. In other words, the position of ngwalda is invariably explainedand justified in terms of the activities performed (Martin 1970, Genest 1976).

The actual performance of these activities, however, is not a necessarycondition; most ngwaldas and ngwalda-lineages have nothing to do withiron-smelting, some have never acted as undertakers, etc.; but the idea thatthe ritual exclusions should not apply to these people would seemincongruous. One is born a member of that group and stays so whateveractivities one performs. In other words, the activities are mentioned as an aposteriori explanation of the exclusion, not as a necessary condition for beinga ngwalda. This idea is suggested by the circular explanations that can befound in Mafa myths. Some of these stories say that ngwaldas were given thedangerous job of undertakers because they were 'strong' enough to handlefurnaces, or alternatively that they had to accept a dangerous activity likeblacksmithing because they were too 'weak' to resist, being polluted bycorpse-handling (Martin 1970: 34ff., Genest 1976: 23).8 Not to put too finea point on it, even when some activities are clearly designated as the exclusivelot of a certain category, the activities are not taken as a necessary conditionfor membership of the category (Boyer 1982b passim).

Conversely, it is often clear that the activity mentioned as a ' definition' ofthe position is a non-sufficient condition, so that one can have the occupationwithout the position. This is precisely what happens with the Fang positionsmentioned above, and it applies to most positions linked to ritual activitiesand more generally to the handling of traditional truths. A Fang mbommvetcan be defined as someone who plays the mvet instrument and knows therepertoire of epic tales. However, many people who fulfill such criteria arenot considered mbommvet at all. This concerns not only young inexperiencedstory-tellers but also musicians coming from distant clans; and it happensthat some persons, formerly said to be mvet players, are not so consideredany longer. All this means that membership of the category mbommvet canbe the subject of conjectures beyond such obvious properties as playing theinstrument and knowing the stories.

In most cases of traditional positions, activities are only part of what isrequired; one has to have ' something more' that marks the ' true' or ' real'holders of such positions. This 'something more' is usually not defined; theFang express it in terms of having 'eaten the medicine' (i.e., undergone thespecific initiation) for this or that activity. This is very vague, and mostpeople have no idea at all how one could 'eat' this 'medicine' or what itconsists of. This in fact suggests that the persons considered are differentfrom ordinary people in that they are evur-bearers. The idea of a 'hidden'property that explains why some mvet-playing persons are not mbommvet,means that the typical activity, however central, is not a sufficient argumentfor ascribing someone a categorical status; this conclusion I think applies tomost traditional positions.

102

Customised persons

The idea that traditional positions are classified in terms of activitiescontains some truth, in that occupations are the idiom in which ideas aboutpositions are usually expressed. However, the representation of activitiescannot be the mental definition of the positions, since it provides neithernecessary nor sufficient conditions for membership of the category. It issometimes assumed that such phenomena (people performing the activitieswithout having the position, or conversely) are just exceptions in theotherwise smooth operation of the social system. People, however, do notthink of these cases as exceptions in a system of rules; they do not considernon-blacksmithing blacksmiths as being outside of the system; such personsare fully-fledged blacksmiths like the others, in much the same way as anorange-coloured lemon is strange but is still a lemon, neither an orange nora 'half-lemon'. I think this allows us to draw some conclusions about the waypeople represent the link between activity and category; these conclusionsapply not only to the so-called exceptions but also to cases where theactivities expected are effectively performed.

As I pointed out above, we are dealing here with named positions, i.e., witha set of categories. The problem is what makes it possible to decide that acertain person is an exemplar of the category. The activities usually linkedwith a certain position can be described as a typical identification of thecategory. The important point here is the idea of typicality, i.e., of usingcriteria which allow one to recognise most exemplars of a category, but areinsufficient to determine the extension of the term. When asked what ambommvet is, Fang people almost always answer that it is someone whoplays the mvet. When reminded that so and so plays the mvet and is notconsidered a mbommvet, they typically answer' that's true, but you see that'sbecause he isn't a mbommvet'. What is clearly implied by the use of thecategory mbommvet is that the people to whom the name applies aredifferent from the rest of the Fang. In what way they are different is notconsidered, but what is known for sure is that they must be different,otherwise they would not be mbommvet. Not to put too fine a point on it,the representation of the position is a yes/no question: one is or is not angengang; people might be unsure in certain cases, but that never means thatone could be 'half-ngengang' or 'mbommvet to a certain extent'. Whatmakes one a mbommvet or ngengang, however, is not such observablefeatures as playing an instrument or organizing certain rituals. It issomething more that all members of those categories have. No one knowswhat it consists of, but it has to be there, otherwise, whatever one's activities,one is not an exemplar of the category. External typical criteria are justindirect (and insufficient) evidence of the fact that people really belong to thecategory.

103


Living kinds and naturalised positions

This type of identification makes some position terms very similar to naturalkind terms, especially to the terms used to refer to living species. As I said inchapter 2, there cannot be a definition for such terms as 'giraffe' or 'oak'.The representation and use of a natural kind term always involves two typesof general assumptions:

(i) some assumptions about the typical features of the exemplars of thekind. Although most of these features are present in most exemplars,they are neither necessary nor sufficient conditions. Thus, giraffes aresupposed to be long-necked and lemons yellow. But short-neckedgiraffes and blueish lemons are still giraffes and lemons, so the featuresare not necessary. They only provide a 'stereotype' of the kind, whichcan be represented in various ways (as a series of features, or with thehelp of mental images, etc.) but never constitutes a definition. The useof a natural kind term, however, requires another general assumption,

(ii) the presumption of an underlying trait that is common to all exemplarsof the kind. This is what makes giraffes giraffes and dogs dogs; thisessential feature is necessary to the use of natural kind terms. Using theterm 'giraffe' implies that beyond surface resemblance, there issomething common to all real giraffes, although what this common traitconsists of is left undefined.9

There is good psychological evidence that both types of assumption arementally represented and govern people's use of such terms, especially in thedomain of living kinds. Even in young children (four-year-olds), thepresumption that living kinds have a common essence is so strong that itoverrides perceptual resemblance. Children extend inductive generalisationsabout an exemplar to other exemplars of the same kind, even if they lookvery different, but never to exemplars of another kind even if they are verysimilar (Gelman and Markman 1987, Gelman 1988).10 The idea of anundefined common essence is a powerful cognitive mechanism, universallyavailable to human minds; together with basic principles of taxonomicordering, it organises biological knowledge in all cultures (Berlin et al. 1973).

It is difficult not to notice the structural similarity between therepresentation of such categories and that of the social positions consideredhere. As I mentioned above, the positions involved in traditional interactionseemed to be related to characteristics such as occupation but the latter donot constitute necessary and sufficient conditions for membership. At thesame time, however, people suppose that there is some underlying featurethat is common to all persons in the position. In other words, there is apresumption that persons occupying a certain position share an essence,although what the essence consists of is left undefined. In many an-

104

Customised persons

thropological accounts, this is considered a puzzling feature. If people aregiven a certain position, there should be a mental' definition' of what makesthem different from others. This paradox is eliminated if we admit thatpositions in such a context are represented in the same way as living kinds.People who are 'mbommvet' or 'ngengang' are thus considered naturallydifferent from others. If this difference is represented in terms of naturalkinds, there is no need of a defining feature; what is necessary is just thepresumption that the essential trait exists, not any hypothesis about what itconsists of. Also, this hypothesis makes it possible to account for the'exceptions' mentioned above. People who are said to belong to a caste ofblacksmiths but never smelt iron are represented in the same way as short-necked giraffes, i.e., as potentially interesting cases but not as non-giraffes. Abeing is or is not a giraffe; there is no way of being ' half a giraffe' or' a giraffeto a certain degree'. Although it may be difficult to identify a concreteexemplar, there is no doubt that it either belongs or does not belong to thekind.

The representation of social positions then relies on the extension to socialdifferences of spontaneous assumptions which prove extremely successful indealing with the natural world.11 This of course applies a fortiori to socialcategories which are explicitly defined as quasi-biological species. The mostsalient illustration of this is the Indian jati which can be glossed as 'caste' inthe social domain and ' species' in the natural domain, but there are countlessother examples. The assumptions of shared essence and typical features seemto be a direct translation, in the domain of traditional interaction, of the wayeveryone from early childhood identifies and classifies natural kinds.

Position and truth

Let me now return to the general question of position and truth. Truth seemsa privilege of certain social positions. Only some actors in traditionalinteraction are supposed to make true statements about certain domains ofreality. This is difficult to understand if we maintain that traditionalstatements are judged on the basis of the 'competence' involved. In such aframework, initiation for instance is just incomprehensible. If, however, weassume that statements are judged on the basis of causal criteria, then it ispossible to interpret such institutions as a way of establishing the causalconnections deemed necessary, or more precisely as a way of making theassumption of such causal links more salient in the listeners' minds. But thisin return makes it necessary to explain why only certain persons should beliable to get involved in these causal processes. My claim is that the only wayof making this assumption is to represent a natural difference between peopleoccupying certain positions and the others. This in fact is a very commonstrategy in both science and common sense: differences in causal power are

105


related to underlying differences in kind. If two animals which apparentlybelong to the same species turn out to breed in significantly different ways,it is obviously relevant to assume that they in fact belong to different species,that the underlying traits are different.

We can now understand the apparent strength and fatal flaws of theordinary authoritarian description of tradition. The basic insight, whichrelates truth to position, is right and faithful to what the actors themselvesreport. The conception is wrong, however, as concerns the representation ofpositions, which is not only a question of classification or ranking, butpresupposes the attribution of a hidden, inscrutable essence to the holders ofcertain positions. Only some people are liable to be affected by a certaincausal process: this idea presupposes an essential difference between thepeople concerned. Indeed such differences are very often expressed asdifferences between species of people, formally comparable to animal species,and justified in terms of biological differences.

Finally, let me mention a possible misunderstanding that should beavoided. This hypothesis of course does not explain why there are socialpositions, not even why there are certain positions and not others, even lesswhy specific people occupy the positions. What it explains, on the otherhand, is why and how, once the positions exist, it is possible for people totake them as a criterion of truth. This, as I said, is impossible to understandif we maintain that positions are mentally defined or that utterances arejudged are the expression of some 'competence'. The connection betweenveracity and specific positions is plausible only if the categories in questionare represented in the same way as living kinds, if the positions arenaturalised.

106

Conclusions and programme

In the course of this book I have left aside some problems which are usuallyconsidered crucial in a characterisation of tradition, concerning, e.g. theclassical opposition of 'tradition' and 'modernity', the question of'traditional societies' and of a 'traditional mentality', the problem of therelationship between literacy and oral traditions, etc. Although theseproblems should all be dealt with in a theory of tradition, it is difficult to gobeyond common platitudes if we do not have a precise description oftraditional interaction itself. In this chapter I will try to re-examine thesequestions in the light of the hypotheses put forward about the intellectualprocesses involved in traditional discourse. I will start with a general pictureof the theory so far, then examine some of its possible implications, andfinally suggest some directions for further research.

We examined here a limited domain of traditional interaction, namely thecognitive processes involved in traditional discourse. This way of ap-proaching the subject is not meant to suggest that traditions are just series ofstatements. Obviously, many aspects of traditional situations are non-verbal;these elements are remembered and repeated and constitute traditionalobjects, often but not always in combination with verbal communication. Inpresenting a mainly ' verbal' account of traditions, I do not mean to implythat non-discursive aspects are secondary; my reason for choosing to focuson utterances was mainly tactical. In the common anthropological views,traditions are described as founded on some set of conserved underlyingpropositions about the world. If this were the case, the analysis of traditionalutterances should not pose too many difficult problems. Obviously, languageis the richest representative idiom in which propositions, 'theories', 'world-views' and 'conceptions' could be expressed, much more easily than indance, ritual gestures, pictorial representations, etc. In other words, iftraditions were about world-views, the study of traditional discourse shouldhave no difficulty in uncovering them.

It turns out, however, that traditional utterances cannot really be linkedto underlying conceptions without systematically distorting the data,overlooking many of their important properties or creating difficult

107


problems. True, the different types of traditional discourse I have examinedcould be described in terms of world-views transmitted, but at the price offocusing on irrelevant aspects and leaving aside the crucial features of eachtype of discourse. For instance, traditional truths in certain societies must beexpressed in a ritual language whose structure minimizes the 'meanings'conveyed. One could still claim that some 'meanings' are transmitted despitethe ritual language. But what then is the point of conveying them in anobscure, misleading form? One would have to interpret this as some form of'noise' in the communication of world-views. As it happens, however, thepeople concerned take the formalisation of speech to be, not a contextual orsecondary property, but a crucial feature, so much so that a translation inordinary language is not supposed to convey a truth at all. By taking theexpression or transmission of information as the main point of the ritual, wewould both create a difficult problem (why do people bother to usecomplicated verbal forms?) and consider irrelevant what the actors take tobe essential, what they often try to transmit from generation to generation.As we have seen, the same point can be made about divination or initiationrites as the origin of true discourse.

This has some consequences in the study of traditions in general. Iftraditional statements do not convey a world-view, then other aspects oftraditions (gestures, dance, and other types of action) must be a fortiori evenless 'expressive' in that sense. The actual study of traditions, as complexphenomena of memorisation which integrate specific gestures and actions aswell as utterances, has not yet been undertaken in anthropology; but onepoint at least is clear: such a study cannot start from the assumption that theexpression of underlying theories or conceptions is the principal aspect oftraditional actions.

Traditions and psychological hypotheses

Theories of tradition necessarily imply some claims about mental represent-ations and processes. Reiterated phenomena could not be reiterated withoutpeople memorising them; salient actions are salient only if people's attentionis mobilised by them; literalism is a psychological attitude toward variationsat the 'surface' of the cultural phenomena considered. As I pointed out in theforeword, it is difficult to put forward a plausible theory of tradition withouthaving some clear psychological assumptions. This applies to commonaccounts of traditional interaction too: terms such as 'conservatism' or'traditionalism' refer to cognitive and emotive states, processes andmechanisms.

The processes described in the previous chapters are for the most partsimple and common properties of people's reasonings. Three features appearto be particularly important in the interpretation of traditional contexts.

108


(i) Causal criteria of truth. The listeners judge a certain utterance truebecause they represent the existence of a causal link between a current stateof affairs and the event of the utterance about it. This of course is an abstractanalytical description. People do not have to represent such logical terms as'causal links', 'states of affairs', etc., to represent a singular causalinterpretation of an event, e.g. an utterance. People's criteria of truth arealways expressed in the concrete terms; a specific state of affairs causes aspecific event, in this case a singular utterance.

(ii) Anchored categories. This refers to the process whereby a notion isstripped down of its general definitional features. Some traditional notionshave no definitional features at all and others are prima facie ' normal' ordefinable notions, about which it is then assumed that the accepteddefinitional features are insufficient for a really referential use of the term.The distinction between common use and truth-making registers is essential.The acquisition of expert discourse implies, not the acquisition of newgeneral representations about the entities designated, but the memorisationof ostensive presentations of these entities or properties.

(iii) Naturalised positions. This is another aspect which is often mentionedin anthropological texts, although its relevance for a general theory oftraditions is not clearly perceived. Positions are conceived in terms ofessence, in much the same way as living kinds. The accession to a certainposition is therefore conceived as the unfolding of underlying propertiesrather than as a contingent change in people's activities, even when such achange is the only means to recognize the position. The representation ofpositions as natural makes it possible to conceive certain causal links asaffecting certain persons only.

The original feature, the characteristic of traditions, is the combination ofthese processes. This combination is found in situations where claims aremade about the truth of certain statements, and about the efficacy of certainrecipes for producing truths. This process obviously involves not only acertain organisation of the utterances, but also a certain type of cooperationon the part of the addressees, and some specific ways of memorising suchsituations. In other words, tradition is conceived here as a specific type ofcommunication, not in the restricted sense of a transmission of information,but rather as a type of interaction which modifies people's representations ina relatively organised way.

These three hypotheses are no doubt insufficient to constitute a completetheory of traditions; there may be other essential features of this type ofcultural institution, and the features mentioned here perhaps deserves a morerefined psychological description. The hypotheses, however, make it possibleto account for some of the general properties with which anthropologists arefamiliar, yet are seldom taken into account in theories. Among suchproperties are the criteria of traditional style I mentioned in chapter 1,

109


namely the fact that people's attention is focused on singular events ratherthan general principles, the consequent literalism and the fact that someinstitutions are supposedly justified by the fact that they have been donebefore in much the same way. Common conceptions of tradition have littleto say about these common properties. They cannot account for theinsistence on singular instances, which is virtually denied in intellectualistmodels: people are supposed to think in theoretical terms, although they areconstantly talking about singular cases. As for literalism, it is interpreted asa consequence of people's traditionalism, but this is certainly no explanation;the idea of a spontaneous conservatism is too vague to be of any explanatoryvalue. The fact that repetition is a sufficient justification for performingcertain actions receives no explanation at all. A theory of tradition shouldcertainly be judged on the type of account it gives for such general properties;the cognitive hypotheses put forward here were chosen precisely because theymake it possible to account for these common aspects of traditionalinstitutions.

Implications (1): traditions and societies

In the course of this argument I have described traditions without mentioningthe anthropological and philosophical debates about the 'Great Divide'between 'tradition' and 'modernity', these being understood as differenttypes of societies or different mentalities. The societies generally considered'traditional' in anthropological dichotomies can be roughly described asexotic, small-scale peasant or pastoral communities. The Fang, Beti, Cuna,Mianmin, Mundang, etc., from whose institutions I drew my examples andillustrations, live in such groups. So do most of the people classicalanthropological monographs are about. There is therefore a strong tendencyin anthropology to think that generalities about traditions are necessarilygeneralities about this type of societies, as opposed to large empires orkingdoms, modern industrial nations or ancient city-states.

In the conception proposed here, the question may be raised, whether thetype of interaction I have described as the foundation of traditionalrepetition can be found only in a certain type of society. ' Great Divide'models just beg these questions, since concepts such as ' traditional mentality'or 'traditional society' are the starting point of the theoretical constructions.This of course is an empirical question, not a matter of a priori judgement.I have mainly described traditions in terms of processes of representationsand communication, which may be typical of a certain type of socialenvironment, or not. I will argue that the latter is more likely, in other wordsthat traditions can be found in very diverse types of societies, although wemust be wary of some conceptual confusions about this question.

This position may seem odd, since all the examples I have mentioned in the110


course of this book were taken from societies which would fall on the'traditional' side of any 'Great Divide' conception. The reason for this is notthat examples could not be found elsewhere, but a simpler point of method.If we want to uncover some general properties of tradition, it is certainlysafer, and I hope more convincing, to take as a starting point institutionswhich are uncontroversial examples of' traditions' in the usual sense. At thesame time however, most of the phenomena I have described here must havea familiar ring to anthropologists who have worked in those supposedly' non-traditional' environments. Divination is rife in industrial societies (andprobably widespread in all civilisations); so is the idea that specificindividuals have a particular relationship to hidden entities or agencies,which allows them to state the truth about certain domains of experience. Totake a more trivial example, it is difficult for an anthropologist to describe the4 bullying' to which neophytes are subjected in many modern institutions,notably the armies, as anything but a form of initiation. Does this mean thatwhat I have proposed here as the general properties of traditions can befound in most, maybe all societies? My answer (necessarily a tentative onesince this is an empirical question) is that traditional interaction can indeedbe found in other societies than prototypical small exotic communities; towhich I hasten to add the reservation, that mere similarities of institutions(such as in the examples above) are certainly not sufficient to infer that oneis dealing with a traditional interaction, with all the properties describedhere. To substantiate what may seem a Byzantine qualification, I will givetwo opposite examples of modern institutions, which seem to me clear casesof non-traditional and traditional interaction respectively.

The first example concerns 'experts' in modern industrial societies, e.g.,specialists of computers or of the economy, who are treated with implicitconfidence as sources of truth about certain domains of reality by a ' generalpublic' who have no means to assess the validity of their arguments in anydetail. This situation may appear, and is often described as, similar to thatof the traditional specialists I have described here. The representationsinvolved, however, are very different, and the resemblance is only asuperficial one. The very idea of competence implies that such experts havein their minds an adequate description of the domain of reality concerned,e.g., the workings of the economy or of the human body. They are alsosupposed to develop an expert 'hunch', a capacity for inferring complexunderlying mechanisms from apparently trivial indices. Such skills areconceived as rooted in learning and practice; the assumption is that thecombination of abstract knowledge and past experience has given thesepeople a view of these domains which corresponds to the reality.

The picture is very different in the case of traditional specialists. If myhypotheses are correct, then the very idea of'competence' as described aboveis irrelevant. In certain people's utterances, truths are expressed because these

111


utterances are made in such situations that they are constrained andprovoked by the realities themselves. The speaker's representations are, as Isaid in chapter 5, entirely by-passed. Elders who talk on behalf of theancestors are in fact letting the ancestors talk through them, and the idea ofreal states of affairs directly influencing a statement is fundamental indivination rituals too. In other words, a traditional specialist is not someonewho has an adequate picture of some reality in her or his mind, but someonewhose utterances can be, in some contexts, directly determined by the realityin question. This crucial difference also explains why expert statements andthe reproduction of expert knowledge are not at all comparable withritualised statements or initiation processes. The fact that people trustexperts implicitly in modern industrial societies does not imply that thecriteria of truth and the process of legitimation are the same.

A strikingly different situation is that of the London ' witches' described inT. Luhrmann's monograph and articles (1986). These, I hasten to say, arenot members of a bizarre sect in search of ecstatic experience. Most of themare sober and rational people, who use the various resources of European orother folklore to create situations in which the interplay of ritual action andfree imaginative associations makes it possible to experience ' another level ofreality', and supposedly to have access to truths beyond the reach ofeveryday cognition. Their view of the effects of magic is rather pedestrian,based on a rather empiricist idea of trial and error rather than 'mystical'adherence to a creed. The interaction that takes place in these 'covens' cancertainly be described as traditional, in the sense that it displays all thefeatures I have described above. The crucial notions are ordinary terms madeundefinable. The specific rhetoric of magical discourse aims at deprivingcommon words of their ordinary reference, so that the 4 forces' and otherelements of magical action are considered beyond definition or description.Acquiring these notions means diverging from their everyday understandingand going through a series of situations which constitute ostensivepresentations, e.g. of the magical 'force' or 'power'. Each group is led by an' adept', whose utterances are supposed to convey important truths about thehidden forces and agencies. The 'adept' has a special access to these forces,and is described as being put under 'pressure' or being 'pushed' by them.The guarantee that these persons make true statements is therefore to befound in the fact that their utterances are caused by the forces they aretalking about. To the participants in the ritual, there is an undefined state ofthe forces which causes the adept's utterances. Thus, the pattern ofinteraction between members of the 'covens' has all the features oftraditional communication. And one of the results of this communicativeinteraction is the repetition or reiteration of specific ritual situations.

My aim in making these remarks is not to deny the obvious differencesbetween groups like the Fang and other societies, with a strikingly different

112


social organisation and intellectual climate, but to replace these differences ina broader context, that of the multiplicity of claims to truth made in anysociety at any period. In the previous chapters I mentioned the Fangtradition of epic singing, their divinatory traditions and the tradition ofancestor-cult. There is no unique access to knowledge, but rather a variety ofinstitutions concerned with different domains of reality. Furthermore, in aFang village people make many claims to truth, which are not linked to atraditional situation in any sense. Competent drivers or people who knowabout distant regions of Cameroon are just supposed to have acquired theright skills and representations from the accumulation of relevant experienceor from some appropriate tuition, and nothing more. This in fact applies tomost domains of everyday activities.

What makes the Fang intellectual climate evidently different from that, sayof the Chinese empire, Greek city-states or suburban London is the fact thatmost, in fact nearly all salient, socially significant and memorable claims totruth are made in the context of traditional interaction. These claims to truthare founded on the idea of a causal link between states of affairs andutterances about them; they imply a naturalised representation of people'spositions and the episodic 'anchoring' of certain crucial notions. In short,most truths beyond the sphere of common sense are stated in traditionalsituations. The difference between this, and a society where a significantnumber of socially salient claims to truth are made in non-traditional settingsand interactions, might well be one of degree and not of nature. Obviously,there may be a difference in nature between these types of societies. I am onlyclaiming that there is no corresponding divide, in ways of defending theveracity of certain statements.

These empirical arguments make it possible to understand more preciselywhy Great Divide models are an obstacle rather than a springboard fortheories of traditional interaction. Most of these models pose difficultproblems, notably the following:

(i) in most formulations, 'tradition' and 'modernity' are treated asessences. It is assumed that there are two kinds of societies or cultures.Obviously, this is question-begging; the question, whether societieswhere traditional interaction is observed constitute a different kind, is anempirical question. It cannot be solved a priori, but only in the light ofdata and theories;

(ii) the distinctions are asymmetrical. As Gellner points out (1974: 150), thedistinction is made, either between the domain of' tradition' and an ill-defined variety of non-traditional societies, or conversely between aprecisely defined type of rational thinking and some vague idea oftradition. In the latter case, tradition is a residual category, which lumpstogether Pygmy hunters and Chinese mandarins, Zande herdsmen and

113


Hindu priests. The fact that this set is rather disparate, to put thingsmildly, is obscured by the assumption that all these people are'traditionalistic', again a supposedly unproblematic description;

(iii) this unbalanced description leads authors to focus on the question of thedemise of tradition and the birth of a modern, scientifically orientedculture. This relies on the assumption that people are naturally'traditionalistic', that they normally stick to past generations' beliefs,until extraordinary circumstances force them to abandon that attitude.Once again, one just begs the question why and how the actors happento find past practice convincing and sufficient.

These might be taken as accidental defects in the current formulations ofthe Great Divide question, but I think the flaws are in fact inherent to anysuch dichotomy. There is simply no reason why the properties of traditionsshould be amenable to a dichotomised description. Traditions are a complexform of interaction; they involve a specific distribution of roles betweenpeople, a specific type of criteria to evaluate their utterances, a specific kindof representation about certain cultural categories, and so on. As I tried tomake clear in the rest of the book, the combination of these elements makesit possible for traditional interaction to occur and result in the repetition ofcommunicative events. Most of the elements are likely to be found in othertypes of situations, perhaps even in prototypical 'modern' contexts. Thereasons why simplistic dichotomies are rife in the study of traditions havemore to do with an historically dated obsession with the difference between'us' and 'them', than with any serious scientific argument.

Implications (2): tradition and literacy

I must now turn to a much more specific question, that of the relationshipbetween the type of communicative interaction I have described here and thepresence or absence of literacy. There is a vast literature on this subject,which I will survey only briefly, to indicate some consequences of the ideathat traditions are not systems of ideas or conceptions but patterns of socialinteraction. The importance of the relationships between oral traditions andliteracy has been emphasised by J. Goody in a series of papers and books(Goody and Watt 1963, Goody 1977, 1986, 1987) the main point of whichwas to go beyond the sterile Great Divide approaches, by focusing on actualcommunication processes rather than abstract characteristics of'tradition'.The existence of literacy has profound consequences on people's represent-ations of the world and on the way they organise and communicate them; inother words, literacy does not only provide a technique of informationcommunication and storage, it has some direct effects on the type ofinformation people make use of. Among other things, the study of the effects

114


of literacy makes it possible to refute some deeply entrenched anthropologicalbeliefs; for instance, the lists and tables of binary oppositions, in whichstructuralist anthropology sees the hallmark of the pense'e sauvage, aretypically literate devices (Goody 1977: passim).

The starting point in these studies is the different properties of oral andwritten communication as such. Obviously, the main difference lies in thestatus of the message, which is externally available in written communicationand can therefore be apprehended independently of the originator, while oralmessages are necessarily embedded in the actual face-to-face interaction ofspeaker and listener(s). An important consequence is what Chafe calls'evidentially', the fact that in an oral message it is impossible to conveysome information about a domain of reality without conveying some idea ofthe speaker's relationship to that knowledge (Chafe 1982). Other conse-quences have to do with the internal coherence of written messages, renderedeasier by the fact that all parts of the message are simultaneously available,and various other stylistic differences (Chafe 1982; Tannen 1985: passim).Furthermore, the cognitive processes involved in understanding andmemorising written and oral texts are quite different (Hildyard and Olson1982).1

There is a complex relationship between these properties of oral andwritten communication and traditional interaction. An obvious question iswhether the existence of literacy precludes the type of interaction I havedescribed here. But we must distinguish two different problems here, that ofthe existence of some traditional interactions in a society, and that of thepreeminence of traditional situations as the sole origin of truths beyondcommon sense and everyday experience. The answer to the first question issimple and empirically evident: there are traditional situations andinteractions in many literate societies. I have mentioned the case of Londoncovens, and most third-world societies illustrate the coexistence of traditionalinteraction with the gradual diffusion of literacy.

The second question is more difficult and historically more important. Canwe say that the introduction of literacy automatically leads to the suppressionof the preeminence of traditional sources of truth? The answer to thisobviously depends on the understanding of the very vague notion of'literacy'. One must be careful to distinguish between two different,sometimes divergent aspects of writing systems, which I would call the data-storage and the frozen speech aspects. Scripts can be used to store and retrievedata, i.e., encoded descriptions of states of affairs, and they can also be usedto represent words and sentences of a natural language. The latter use is sofamiliar to us that it is sometimes difficult to take the other aspect intoaccount, and realise that most scripts were either entirely designed for datastorage or used largely for that purpose, to the exclusion of speechrepresentation. For instance, most pictographic systems and mnemonic

115


devices are clearly intended for data storage and retrieval. And many scripts,like, e.g., the ancient Mesopotamian and Mycenean syllabaries, were mainlyused for book-keeping purposes, again to store data rather than freezeutterances in a natural language.2

The likely effects of literacy on traditional interaction certainly depend, toa large extent, on the relative importance of these two aspects in the socialuse of literacy. Scripts used mainly for data storage are by nature outside thedomain of the legitimation of claims to truth; scripts mainly used torepresent speech, on the other hand, can take over from oral communicationin the expression and diffusion of arguments about such claims. Once writingis used in that domain, it is likely to lead to the demise of traditionallegitimation as described here. The idea, for instance, that true utterances arespeech-events directly caused by some real state of affairs, is clearlyinapplicable to written messages, which are necessarily detached fromparticular speakers and circumstances.3 Other properties of traditionalinteraction, such as the strict, quasi-biological identification of truthfulpeople as directly influenced by specific states of affairs, become irrelevanttoo; and this can lead to extreme situations such as that of classical Greece,where very few domains of truth are beyond the written argumentation of theliterate elite.4

Obviously, the 'frozen speech' use of literacy is a necessary but non-sufficient condition for that process. As G. Lloyd points out, the developmentof empirical, polemically ' anti-magical' types of argumentation in classicalGreece cannot be attributed to the sole effects of literacy, which appearedmore than two centuries before; moreover, the division between written andoral does not exactly map that between traditional and empirical arguments(Lloyd 1979: 227, 239ff, see also Scribner and Cole 1981). Even in Greekcity-states, the intellectual climate of which was clearly 'agnostic', literatecommunication was not immediately used for the expression of claims totruth. One can therefore see literacy as a 'multiplier', in the economic sense,rather than the sole determinant of such processes. Casting traditionallycreated truths in the written format is the first step on a slippery slope thatmay lead to the discussion and contestation of these truths.

Truth, acquisition and salience

Let me now turn to possible further directions. I shall begin with a numberof platitudes, in relation to the conclusions of chapter 2. Ideally, ananthropological theory should at least provide some means of understandingwhat makes certain ideas and values so natural and obvious to people livingin a certain cultural environment. Explaining this is as difficult as it isnecessary; not surprisingly, the problem is often ignored or treated with agood measure of handwaving. For instance, people are said to acquire by

116


'socialisation' the 'cultural models' and 'symbols' that give 'meaning' tovarious events. The explanatory power of such ideas is rather small, to putit mildly. Even if given more substance, they imply several not-too-plausiblehypotheses, among which the following:

1 people live in fairly coherent cultural environments, where propositionsabout such things as ancestors or gods are clearly stated, neither toonumerous nor too complex;

2 acquiring such ideas is a simple process. Subjects just pick up what is inthe air, as it were, and use it;

3 there are no strong cognitive constraints on what cultures consist of. Aslong as the 'models' provide 'meaning' where needed, indefinitelyvariable ideas will do.

Unfortunately, each of these hypotheses just flies in the face of the facts.First and foremost, people do not live in such simple cultural environments.People have access to lots of singular actions and utterances, many of whichare unrelated or inconsistent. They must build on these singular data by aprocess of inductive generalisation. The resulting models are not transmitted;what are transmitted, on the other hand, are specific recipes, words, gesturesand other such 'surface' phenomena. Moreover, psychologists know that theacquisition of simple notions or relations is not a simple process; if acquiringthe meaning of 'chair', 'uncle' and 'cause' involves complex mechanisms,this should hold a fortiori for concepts and propositions which are lessobviously anchored in everyday experience. By and large, most cognitivepsychologists agree that the only way of accounting for learning processes isto posit strong innate mechanisms, which produce inductive generalisations,make it possible to weigh their plausibility and correct them, and so on.Inevitably, this implies that there are strong constraints on ' learnability';human subjects cannot form just any hypothesis, cannot learn just anyartificial language, etc. But this of course makes claim 3 rather difficult tomaintain; anything will just not do as a 'cultural model'. There are somelimits to what people can acquire in order to generate 'meaning'.

Going further than these platitudes, however, is especially difficult inanthropology. It would be highly unrealistic to claim that the hypothesesmade in the course of this book provide a comprehensive alternative to theinadequate 'explanations' criticised here. I would claim, however, that theymake it possible to imagine some directions of research, toward a moreempirically plausible account of the acquisition of cultural ideas. In order tomake this rather ambitious claim less abstract, let me return to the centralquestion of the criteria of truth used in traditional interaction. In chapters 2to 6,1 tried to show that what makes certain types of discourse ' traditional'can be described as a specific organisation of the claims to truth? Criteria oftruth are in principle as varied as utterances themselves. Most societies,

117


however, have some institutionalised sources of true statements, with specificcriteria. Certain recipes are supposed to result in the production of truestatements; I have described traditional institutions as such recipes, positingthat some interesting common properties make them different from theclaims to truth which can be found in other situations.

The very notion of a claim to truth implies that the actors do notnecessarily take to be true whatever statement is made according to therecipe. In many cases there is some competition about which recipes are themost appropriate. Even when subjects take a recipe as really effective, theircommitment to the end result is extremely variable, and depends only partlyon the characteristics of the recipe. This is why I tried to keep separate, inchapter 3, the public phenomenon of truth-ascription, about which I have putforward a specific hypothesis, and that of conviction or adherence, whichdeserves a separate study. Truth-ascription is the process whereby a selectionis made, between the utterances which express a truth and those which donot. The outcome of this process is both simple (an utterance conveys sometruth or not) and easily explicated. People's conviction or commitment onthe other hand, their adherence to the proposition expressed by a certainsentence, is a 'private' reality. It is neither simple nor easily described; thereare many shades and degrees of commitment.

In anthropological 'theories of belief, the two phenomena are usuallyconflated, or discussed in the same breath, which suggests that they arenecessarily connected. In such a framework, people find true whatever theyare convinced of, and they are convinced of whatever they say is true. Theproblem with this conception is that it does not provide any means todescribe the varying degree of commitment to a proposition judged ' true';this is problematic, since these nuances of commitment might be extremelyrelevant in the description of some traditional 'beliefs'. More precisely,judging a statement' true' can result in a variety of attitudes, from a slightlydoubtful adherence to complete, rock-bottom commitment. This cannot bedescribed in a theory which recognises only two states, belief and unbelief,corresponding to truth and falsity.

There are no degrees of veracity, but there are degrees of adherence.6 Thefact that subjects can have a varying commitment to propositions which areall marked as ' true' is more easily understood if we take into account thatcriteria of truth are specific representations about an utterance, a speaker, acontext, etc. More precisely, they are assumptions about these realities. Thelistener who judges that the assertion 'ant-hills indicate witchcraft' is truemakes this judgement on the basis of assumptions about the speaker (he hasundergone the witch-experts' initiation), the situation (the statement is madeduring a specific ritual), and other background factors (other statementshave been made about such ant-hills, or about witchcraft being mademanifest by other indices, etc.).

118


Without entering into the complex reasonings involved, we must observethat such assumptions, like all assumptions, have a variable strength orsalience. To take but one aspect of this specific example, the fact that thespeaker is a witchcraft-expert who has undergone the appropriate initiationis not a simple matter. It is an empirical assumption, which can be given moreor less salience by many factors and experiences, by memories of otherpeople's statements, by memories of the speaker's previous utterances, etc.There is simply no way of explaining why some people's utterances arereceived with more credence than others', if we do not consider what makesit salient that they do occupy the position in question. This is why I tried inchapters 3, 4 and 6 to debunk the common idea that people believe a certaindivinatory statement because they hold a certain cultural axiom, which saysthat 'whatever diviners say is true'. Besides its distinct tautological aspect,this explanation is flawed in that it suggests that the identification of expertsis a simple process of classification. In fact, it implies complex processes ofhypothesis formation and confirmation, e.g., about the fact that the personconsidered possesses certain underlying features that make him or her similarto other members of the category, and about the activities as a typical featureof the category, etc.

The way inferences are influenced by the relative salience of assumptionsis extremely difficult to describe, and there is no satisfactory account of thesemechanisms. But the important point here is that they do vary in cognitivesalience, and consequently in the extent to which they play a part in theprocess whereby an utterance is judged true. Cognitive salience is notoriouslydifficult to describe (and even to define precisely), but psychologists workingon everyday, non-demonstrative inference and the role of backgroundknowledge simply have to take this dimension into account.

In order to fulfil the general anthropological brief, and explain what makesit natural to take certain propositions as natural and obvious, we musttherefore describe the processes whereby certain singular assumptionsbecome more salient. This is, in essence, a question of acquisition. It mayseem an impossible task: studying such 'central processes' as belief-fixationis difficult enough in psychology,7 but the difficulty is compounded by thekind of naturalistic data anthropologists have to rely on.8 Traditions,however, may be a privileged domain for such inquiries, in the sense thatmost claims to truth are associated with ritualised situations. Such 'scripted'contexts are the basis for inductive generalisations, as we saw in chapter 2.Their main characteristic is that they make use of drastically restrictedrecurrent material, thereby making it less tantalising to study the evolutionof representations in subjects who undergo and perform such rituals.Notwithstanding the inherent difficulties of such an enterprise, the study oftraditional rituals is bound to be the domain from which theories of culturalacquisition can be elaborated and tested.

119


Conclusion

The main question a theory of tradition should address is that of therepetition or reiteration of specific communicative situations. There is nosatisfactory answer to this question in the anthropological literature, mainlybecause the very problem is obscured by vague notions and conceptualconfusions. Instead of the repetition of concrete situations, most theoriesfocus on an altogether different question, that of the transmission of what isvaguely described as 'conceptions' and 'world-views'. As I argued in chapter1, such intellectual objects may well exist, but they are not of any descriptiveor explanatory value in the interpretation of traditional interaction.

An empirically significant theory of tradition should explain why somespecific situations acquire a special psychological salience, how they areinterpreted in different ways by the different participants, and how thesedifferent processes of memorisation contribute to the reiteration or repetitionof the interaction considered. In this book, obviously, I have not given sucha complete theory. I have tried to give it some foundation, however, by layingstress on the cognitive processes involved in the representation of traditionalcontexts. Concept acquisition, the naturalisation of positions and the use ofcausal criteria of truth are all part of the mechanisms which contribute to thecognitive salience of certain communicative events. It remains to explain howthese elements are integrated in people's representations of the events. Theaim of this book was not to solve the question of traditional repetition, butto build a framework in which it is possible to address it as a significantquestion, as a matter for empirical investigation.

120

Notes

Conserved world-views of salient memories1 Obviously, this distinction between observational and theoretical terms cannot be

taken too strictly, first because 'repetition', like other empirical terms in socialsciences does include some commonly agreed theoretical premises, second,because the distinction between observational and theoretical vocabularies isoften problematic. The claim here is in fact rather modest, namely that therepetition of interaction is among those facts anthropology should explain, whilethe ' conservation of cultural models' is one of the hypotheses the discipline putsforward.

2 Holy and Stuchlik (1983) give a detailed and rigorous account of all theseambiguities, and of the possible options. I have left the question aside because itis neutral as concerns the problem at hand. Whether' world-views' are consideredas being in the actors' heads or analytical constructions does not change the factthat traditional behaviour cannot be described in terms of 'worlds-views', as Iwill try to show below.

3 This is a very brief summary of Horton's views, which cannot be discussed atlength in the space of this chapter. See Boyer (1987) for a more detailedexamination and criticism.

4 Advocates of the ' neo-intellectualist' approach would of course argue that this isnot a definitive argument against the existence of traditional 'theories'. What weneed here is just a 'looser' concept of theory, for which the requirements ofexplicitness and consistency are not so rigid. This, however, will not do, becauseit means trivialising the meaning of the word 'theory' to such an extent that' having a theory' about something is nothing more than having ' some ideas' or' thoughts' about it. In any case, there are more important arguments against theneo-intellectualist view of tradition, namely that it is based on an imaginaryversion of human cognition, and that it makes most traditional phenomenatotally puzzling. I will return to this question in the following chapters. If weconsider traditions as the expression of theories, then it is impossible tounderstand why some traditional categories cannot be defined, why rituals makeconstant use of obscure speech-forms, why divination is convincing. In otherwords, the theory is neither psychologically plausible nor anthropologicallyexplanatory; this I suppose is enough to justify the search for an alternative.

5 The following section is a very sketchy description of the complex network ofrepresentations and beliefs about Fang epic poetry. Boyer (1981, 1982) gives amore detailed account and some information about the ethnographic back-ground.

6 This, obviously, is only a very brief summary of the argument, which is developed

121

Notes to pages 9-24

at length in Boyer 1988, especially about recent changes, between Zwe Nguemain the 1950s and mvet singing in the 1980s. Tessmann (1913) is a very detailedethnographic study of all aspects of Fang culture, which gives some indicationsabout the importance of clan warfare at the time.

7 The analogy is pervasive in anthropology, and the preservation and modificationof cultural traits is often seen as a cultural 'selection' of ideas, formally analogousto the selection of individuals in Darwinism. On this, and the ambiguities of thiscommon idea, see Ingold 1986. I will return to this analogy in the section dealingwith automatic versus intentional repetition, see pp. 13-20.

8 All these problems, to do with the inference of notions from specific statementsand actions, are examined in great detail by Holy and Stuchlik (1983: 55-80),whose general outlook is more optimistic than mine, as they suggest that theproblems are mainly methodological. Also, Holy and Stuchlik do not seem toconsider the difference between representing an event and representing a generaldescription. For instance, they assume that 'all actions are guided by relevantknowledge' (1983: 68) and presuppose that such knowledge consists in generalrules and models. They therefore do not consider the possibility, which I willexplore in the following chapters, that the 'relevant knowledge' in questionconsists of memories of previous actions, with no 'theoretical' principles toorganise them.

9 An additional argument for this is that some aspects of interaction are repeatedwithout being represented at all by the actors. People cannot be conservativeabout what they do not represent. This point will be examined below, in thediscussion of non-intentional models of repetition.

10 This commonsense view of human memory as a bucket filled with pictures andwords is now reinforced by computer analogies. What are called 'memories' incomputers do consist of devices which just store what they are fed and feed it backin the same form. Needless to say, this is a very bad metaphor for human memory,which is a processing and integrating device as much as a storing mechanism.

11 There is a third alternative, that of a Durkheimian 'collective memory' (seeHalbwachs 1925, and Douglas 1982). Although the conceptual frameworks areentirely different, the ideas put forward in this section are consistent with at leastone of Halbwachs points, namely that social interaction need not be totallyrepresented to be reproduced. The idea of'collective memory', however, entailsa ' superorganic' view of culture which is precisely what I am trying to avoid inthe characterisation of tradition.

12 Some psychological conditions of memory for stories were first described inBartlett's famous studies (1932: see also Bartlett 1923 and 1958 for the generalframework). The hypotheses presented here are often inspired by Bartlett's resultsand speculative claims. For a more recent and detailed study of structuralconstraints on stories (and a survey of the literature) see Beaugrande 1982.


1 I will not discuss here the Wittgensteinian or 'linguistic philosophy' idea thatconceptual analysis is quite sufficient, and that in fact it is all there is to do in orderto describe theories and conceptions. Although such ideas once had someinfluence in anthropology (see Gellner 1974, 1979 for a survey and criticism), thepsychological implications are just too implausible to deserve a lengthy discussion.The rest of this chapter in fact provides some arguments against such inferencesfrom linguistic use to theoretical premises.

122

Notes to pages 25-38

2 The following is a very brief description of the ideas associated with evur. Adetailed description of the rituals concerned with evur-related diseases, and ageneral expose of the system, can be found in Mallart-Guimera 1975, 1981. Amore limited description, to do with evur and epic singing, can be found in Boyer1988. On the notion itself and the general anthropological problem, see Boyer1986, 1989a.

3 For other examples of such notions, see the essays gathered in Fogelson andAdams (ed., 1977), although the hypotheses presented in this collection are notentirely satisfactory. Most of the contributors think that the notions in questionsare concepts of'power' and take this gloss as an unproblematic translation. Also,they all consider that the analysis should focus on the 'cultural theories' aboutsuch concepts, a claim which I discuss in the following pages.

4 Keesing's paper is also important in the sense that it shows how anthropology hasa built-in tendency to see 'cultural metaphysics' in local categories, which in somecases are used as simple conventional metaphors. I will return to this point inchapter 3 (see pp. 48ff, and note 4). Keesing 1985 gives a more detailed criticismof 'anthropological metaphysics' and expands on the notion of conventionalmetaphors. For a discussion of the relevance of this concept, see Boyer 1989a.

5 This of course is a simplified version of a rather complicated text, the formulationof which makes the discussion difficult.

6 This criticism applies to many anthropological constructions, besides thestructuralist idea that traditional categories can be construed as 'empty signiflers'.The idea that language embodies or carried cultural 'theories' is pervasive inanthropology, and the argument of this chapter is precisely to show how thisassumption precludes an empirical study of the way categories are actually used.

7 This is the only way of making plausible that Fang people have ' theories' aboutevur. But the claim then is trivial. No-one would deny that Fang people havethoughts about evur, or that such thoughts are organised to some extent. But thenthere is no form of thinking or cognitive activity that is not ' theoretical' in thatsense, so that the term is useless, as I mentioned above (see chapter 1, note 4).

8 The question of the type of' knowledge' transmitted in initiation rites, and of theother cognitive effects of initiation, is discussed in more detail in the first sectionsof chapter 6.

9 For the sake of argument, I admit here that some concepts may be representedwith a mental definition. If on the other hand we accept, like many linguists andmost psychologists, that virtually no terms can be defined and that the resourcesof the common vocabulary are more or less equivalent to conceptual resources,then this holds a fortiori for traditional categories. On the idea that definitions arenot 'psychologically real', see J. D. Fodor et al. (1975), J. A. Fodor et al. (1980).

10 On the notion of natural kinds, a basic reference is Schwarz (1979), with a simpledistinction between 'natural' and 'nominal' kinds. See also Churchland (1985),and Schwarz (ed., 1977), for various contributions on natural kinds and linguisticreference. The representation of natural kinds and the parallel with sometraditional categories are discussed again in the second part of chapter 6, abouttraditional positions (p. 10Iff). See also Boyer 1989a for a general discussion ofpseudo-natural kinds in traditional interactions.

11 A classical source on the various theories of concept representation is Smith andMedin (1981). The view of natural kinds presented here is clearly inspired byPutnam's account (1975, 1983). However, the claims made here are neutral asconcerns controversies on the format of conceptual representations. Whatever theexact format of stereotypes, they are not what one acquires through initiation ritesor other traditional situations.

123


12 Obviously, the problem is to describe the mechanism whereby people pick outcertain common features of exemplars and discard others in the construction ofa stereotype. For general surveys of theories and arguments about this abstractionprocess, see Posner and Keele (1968), Carey (1978, 1982, 1985) and MacNamara(1982).

13 This hypothesis is defended in Boyer (1986), where 'evur' and such terms as'mana', 'mangu', 'orenda', etc., are construed as essentially similar. Thishowever is refuted, partly by empirical arguments (see Keesing 1984, for instance,for a better analysis of 'mana', and Boyer 1989a for comments), and partly fortheoretical reasons explained in the following pages. The interesting aspect oftraditional notions concerns the process whereby they are anchored to specificexperiences, and this can apply to virtually any kind of concepts.

14 The ancestor-cults, in which the notion of bekong is crucial, are described in detailin Fernandez (1982). See also Mallart-Guimera (1981) for the connectionsbetween evur and bekong.

15 See Tulving (1983 passim) for a presentation of the distinction and the casefor the existence of separate mechanisms. The reasons why some psychologistsare reluctant are (i) the fact that certain memorised elements, like e.g. scriptedactions, seem neither 'episodic' nor 'semantic' in essence, and (ii) the fact that itmay be wrong to infer, from different types of memorised material (episodes vs.general propositions) to different devices (see Anderson and Ross 1980). Theargument developed here only implies that episodic material exists and isprocessed as such, a fact that no psychologist seems to deny, yet is obviouslyignored in anthropological theories.

Criteria of truth1 See Luther (1935) and Detienne (1966) for symmetrical, perhaps complementary

interpretations of the evolution of aletheia in Archaic times. While Luther seeksto describe the evolution from mythical to philosophical conceptions of truth,Detienne treats aletheia as a term of the vocabulary, the usage of which can bedescribed in terms of oppositions and implications. My argument on truth-termsis partly inspired by Detienne's stimulating account, although he does not base hisargument on the idea that logical truth predicates are always present, whateverthe metaphorical terms which denote them.

2 Obviously, Gutmann's argument, like all arguments from etymology, is dubiousin that it presupposes that people's use of a term is influenced by the term'sderivation. But not all arguments on 'specific concepts of truth', different fromlogical concepts, are based on such etymological inferences, so I will not discussthis specific flaw.

3 This of course is a postulate. The Aristotelian characterisation is the simplest andthe clearest one available. Also, it has the advantage of highlighting the' redundancy' of truth-ascriptions, a point to which I will return presently.

4 Keesing (1984, 1985) gives a clear exposition of Lakoffs argument and itsanthropological implications, especially against 'anthropological metaphysics', apoint I mentioned in chapter 2 (see p. 29ff.).

5 Obviously, there are many more arguments against the idea that some people donot possess the truth predicates in their conceptual equipment. For instance, itwould be difficult for small children to play (e.g., to pretend they are mice anddragons) if they did not have a concept that includes the logical predicates of truthand falsity. Also, they would have difficulties learning to lie. More generally,

124


truth-terms in all languages pass the test of redundancy which is typical of truthpredicates. If I say that "'snow is white' is abele", this sentence too is abele. Onlytruth-predicates can have this effect, which is the basis of redundancy accounts oftruth predicates; asserting that a sentence is true has the same value as assertingthe sentence itself (see, e.g., Strawson 1949).

6 The existence of'covert categories' (i.e., unnamed concepts in folk taxonomies) isa salient example of the complex relationships between vocabulary and mentalconcepts. An even simpler illustration is given by the existence of ontologicalcategories (about what types of things there are in the world) which are notnamed. In Keil's experiments for instance (1979, 1986), children make inferencesabout new terms on the assumption that they design living things rather thanartefacts, although their vocabulary does not include terms like 'living' or'artefact'.

7 The idea of explaining strange traditional truths by resorting to specificmodalities of belief is in fact so entrenched that it seems to resist even the mostclearly and carefully argued exposition of the ambiguities and problems generatedby the very notion of 'belief. Needham (1972) is a good example of such anargument; although its results should be devastating for most theories based onthe notion of belief, it is simply ignored. I will make another attempt here (a sideattack, as it were) to show that the notion, even if it was clear, would be useless.

Customised speech (I): truth without intentions1 For instance, the English Poor Robin, a seventeenth-century parody of the then

popular astrological almanacs, includes the following prediction: 'we may expectsome showers or rain either this month or the next, or the next after that, or elsewe shall have a very dry spring' (quoted by Thomas, 1971: 398).

2 This is all the more surprising, as many anthropological arguments aboutdivination were put forward in the context of abstract discussions about the'rationality' of divination procedures. As a result, they tended to emphasise thecognitive dimensions of divination as opposed to its practical purposes. This ofcourse is only part of the story. Many divinatory statements concern current, notfuture states of affairs; they relate them to non-observable entities, e.g., gods orancestors, instead of making claims about their subsequent evolution. Moreover,divination is often more directed to decision-making than to understanding theworld, so that 'prediction' is replaced with directives.

3 This description is of course truncated, since people's representations of theactual problem at hand (e.g., bad crops or disease) are left aside. The reason fornot mentioning them here is that I am trying to describe what is specific to thedivinatory diagnosis, as opposed to other statements that could be made on thecurrent problem. I will therefore focus on these elements (the procedure and thelink established between undefined ' situation' and ' diagnosis') which would notbe present in other types of statements.

4 Cicero himself makes an analogous distinction, in Quintus's plea for divination(De Div. I.vi.12), between 'natural' and 'artificial', and then between differenttypes of artificial divination. The 'natural' type belongs to what anthropologistswould rather call 'inductive' divination, i.e., the observation of natural portents(clouds, etc.) and the extension of these principles to other domains.

5 The directly intentional interpretation (following which the oracle is a directmessage from the god) was of course disputed, but only because it seemed toimply that the god was at the client's disposal. Plutarch for instance is quite

125


definite that it is unreasonable to say that the god himself has composed thepoem; it is therefore unreasonable to find the oracles less reliable when they areexpressed in prose. The god only gives the 'original stimulation' (Pyth. Orac.397c). An alternative hypothesis was that the oracles were triggered, not by thegod but by less worthy 'daemons', a solution which made it possible to salvagethe god's transcendence, while maintaining the idea of divination as com-munication.

6 There is another reason for the apparent plausibility of the semiotic interpretation,which is specific to Western culture. Ever since the Renaissance, theories about'correspondences' and 'signatures' have been diffused and influential. The ideathat there are hidden connections in natural phenomena, and that some links ofresemblance connect objects from different realms, was a major tenet of theRenaissance return to Platonism, and I think the influence of that model may stillbe at work in some anthropological models. The idea of signatures, which is moststrongly argued by such influential authors as Paracelsus, was subsequentlydiscarded by scientific inquiry. Precisely for that reason, however, I suppose itremains the model Westerners spontaneously build, when dealing with 'occult'knowledge and activities; it may also be the model people engaged in suchactivities, even nowadays, try to follow.

7 For a good illustration of this, see Humphrey's study of Buryat omens (1976).The Buryat single out many natural events as ominous. Although some recurrentfeatures may be found in the events thus highlighted, there is no systematiccorrespondence between omen and explanation. Rather, explanations are 'free-floating', as Humphrey puts it, so that they are selected by virtue of theirrelevance to the problem at hand.

8 This of course is extremely short, and cannot be substantiated in the space of thischapter, which is more focused on divination in the usual, 'instrumental' or'inspired' sense than on natural portents.

9 See Boyer (1989b) for more details on common anthropological ways ofexplaining strange causal inferences.

10 Here I am only giving the conclusions of arguments explained elsewhere (Boyer1989b), and I cannot but give a few references to substantiate the main point. Theinability of philosophical analyses of 'cause' to explain everyday causaljudgements is illustrated in Hart and Honore's famous study on Causation in theLaw (1959). A classical source for conceptions of causality is Mackie (1974). Asfor cognitive research on causal inferences, see Laurendau and Pinard (1962) fora classical Piagetian framework and Carey (1985) for its refutation. Perception ofcore cases of causal connection by infants is explored in Leslie (1979) and the roleof background knowledge is examined in Schustack and Stern berg (1981) amongothers.

11 The 'photograph analogy' is often used in theories of linguistic reference,especially in 'historical' theories, following which a term like 'giraffe' refers togiraffes if there is a historical (causal) connection between the use of the term andactual tokens of the species 'giraffe'. See Evans (1982: 76-9 and 8 Iff.) for a surveyof these theories and the ambiguity of the photograph model. Needless to say, Iam not attributing this theory to diviners and their clients. The hypothesis putforward about divination only implies that people consider certain singularspeech events as caused by certain singular situations. Such assumptions preciselydo not constitute a ' theory' of truth or reference, they provide specific criteria oftruth.

12 This distinction is explained in more detail in chapter 5, in the examination oficonicity in ritual languages.126


Customised speech (II): truth without meaning

1 For a general description of 'special languages', see Halliday 1976. Althoughsecret codes and languages are extremely common, there is no extensive study oftheir form and usage. About the use of secret codes in initiation rites, see Boyer1980, especially about the relationship between the choice of certain formalprocedures in the creation of the code on the one hand, and the status of thegroups concerned on the other.

2 Sherzer is quite definite on the fact that the ordinary language translations areseen as useful helps for non-specialists, which however do not carry the sameguarantees of truth as the original. On this and other aspects of the specialiseddiscourse, see Severi 1985, and forthcoming.

3 Obviously, this is a recurrent problem in anthropological theories which focus onthe ' conceptions' allegedly expressed in traditional contexts. In the study of oralliterature, for instance, one is often faced with the same dilemma. If we analysestories in terms of underlying cultural 'theories', we focus on aspects which, ifthey are there at all, cannot be relevant for the listeners. So it becomes difficult toexplain why the stories are actually told and memorised. If, on the other hand, wewant to explain this, then we must focus on the mechanisms whereby the storiescommunicate new ideas or reorganise old ones, in other words, on what is not priorcommon knowledge. On this precise point, see Boyer 1982a, 1984, 1987.

4 For a clear exposition of information in 'information theory' (and of the possiblemisinterpretations), see Dretske 1981, which also deals with the relationshipbetween 'information' in this narrow sense, on the one hand, and linguisticmeaning, on the other.

5 Obviously, Bloch's is an 'asymptotic' argument. In actual fact, formalisationonly tends to eliminate choice, and therefore tends to force listeners into thedilemma of either accepting the whole discourse or else dissenting. In thediscussion I will take this into account, and the arguments are supposed to showthat even as a hypothesis about 'tendencies', it is insufficient.

6 See the beginning of chapter 4. The obsessive emphasis on the fact that divinerscan 'manipulate' their statements implies that the procedure is, to some extent,an obstacle to the expert's insightful diagnoses.

7 Although I only discuss Bloch's hypotheses, they are representative of ananthropological tendency to interpret 'efficient' speech and all ritualised speechforms in terms of speech acts. The idea of applying speech acts models to magicalutterances (spells, for instance) was first put forward by Tambiah (1968) andFinnegan (1969). The main idea is that people who recite magical incantations arenot making claims about a causal connection between the words and thesubsequent effects. They are in fact making a special speech-act, comparable towishing, swearing, cursing, etc. Ahern (1979, 1982) is an example of such a speechact explanation of magical action. Gardner (1983) presents a general criticism ofthese attempts, which rely on a rather distorted view of the intellectualmechanisms involved, and only account for marginal aspects of the rituals. Inchapter 5, I present a picture of causal thinking which I think makes it possibleto understand that magical claims are causal claims, without necessarily implyinga strange conception of causation.

The criticisms against Bloch's formulation of course apply also to Finneganand Tambiah, who seem to take the phrase 'illocutionary force' literally, asdenoting some form of energy. I have chosen to insist on Bloch's formulationbecause it is subtler, and also because Bloch tries to give some explanation of the127


idea that more formalisation seems to result in more efficacy, while the otherauthors take that as almost self-evident.

8 The clearest account of' illocutionary force' in various pragmatic theories can befound in Levinson (1983). On the question of the relationship betweenpropositional structure and illocution, see Katz 1977.

9 To be fair, Bloch's formulation suggests that there is a distinction betweenillocutionary and perlocutionary aspects of discourse. But the main hypothesis isthat, in ritual speech at least, the illocutionary force of certain utterances is suchthat certain effects (notably persuasion) are inevitable. In short, certain typicalperlocutionary effects are directly determined by illocutionary properties. This iswhere Bloch's account is consonant with the magic-as-speech-acts schoolmentioned above (see this chapter, note 7), which claims that the alleged ' effects'of spells are a simple consequence of their illocutionary properties.

10 For other examples of such phenomena, see, e.g., Jakobson's remarks on Russianparallelism and glossolalia (1966, 1981). The hypothesis on iconicity presentedhere is of course inspired by Jakobson's ideas, although his notion of ' poeticfunction' does not have much explanatory power as concerns the cognitive effectsof such manipulations of language. See Kiparsky (1976) for a treatment ofidioms, formulaic speech and, more generally, ' fixed' expressions in the contextof generative-transformational grammars.

11 What I have presented here is a very abstract description; an empiricalinvestigation should describe more precisely in what manner the existence ofcausal criteria of truth results in the repetition of traditional events. I will returnto this crucial point in chapter 7.

Customised persons: initiation, competence and position1 I am aware of the vagueness of this general characterisation, although I would

claim that these features are really general, and the necessary starting point of atheory of initiations. A more detailed account, which is more or less compatiblewith what I say in the following sections, can be found in J. La Fontaine's generalsurvey (1985). For the sake of clarity I only consider the case of collective (' tribal')initiations, although what I say about them applies, mutatis mutandis, to personalinitiations as well. See Boyer 1980, 1984 for the analysis of both types ofinstitutions.

2 There is no space here to comment on the cognitive effects of such deceptive'revelations'. A typical, yet paradoxical effect is that they seem to reinforce theneophytes' conviction that the rite is indeed important, that there is more thanmeets the eye in ritual procedures, which cannot be reduced to the farcicalstaging. The mechanisms of this effect are not entirely clear, though. See Barth1975 for a detailed exposition, Boyer 1980 and Houseman 1984 for a discussion.

3 This of course is directly related to the fact that ritual idioms, though they aresupposed to convey important truths, have very little ' propositional potential';see the discussion in chapter 5.

4 Herdt's collection on male initiation in Papua New Guinea (1982) gives a goodsurvey of particularly paradoxical forms of initiation, in which ' knowledge' andpainful ordeals are intimately connected. The essays in this volume also focus onthe emotive aspects of the ordeals, which are obviously crucial.

5 For a characterisation of positions in terms of rights and duties, see for instanceBendix and Lipset (1966), or Goodenough's 'componential' analysis of statusesand roles (1965). Positions in that framework are abstract combinations, which arenot necessarily represented by the actors. Goodenough makes a clear distinction128


between objective statuses and represented 'social categories'. Only the latter isthe subject matter of my argument.

6 I am not denying here that the distribution of positions can be said to constitutea system. But this idea often relies on the assumption that positions are mentallydefined, which I think is not true in the case of the positions implied in traditionalinteraction.

7 There is an extensive literature on these low-status endogamous groups and thesymbolic justification of their exclusion. See Tuden and Plotnicov's collection(1970) for a general survey, especially Vaughan's essay. Whether such groups canbe called ' castes' or not is of course a terminological more than a substantialquestion (Todd 1977). They have all the features associated with castes in theusual, non- or pre-Dumont use of the term (Berreman 1968). One can apply tosuch groups the general claims made by Hocart about the connection betweenposition and social categorising; this is argued in Boyer (1982b).

8 These stories are compared and analysed in Boyer (1982b), with a contrast withnon-endogamous, high-status blacksmith lineages in Central Africa.

9 This of course is a drastically shortened version of the processes involved in therepresentation of natural kinds, especially living kinds. See the references inchapter 2, note 10.

10 See also Keil 1986. These experiments, and their anthropological consequences,are discussed in Boyer 1989a, where it is argued that most complex culturalcategories are linked to the identification of 'pseudo-kinds'. The notionstherefore carry no ' theory' of the entities designated, only the presumption thatsuch entities share essential properties in the same way as living kinds. Theconnection, between representing 'mystical' entities as members of living kinds,and ascribing to them specific causal features, is discussed in Boyer 1989b.

11 The claim here concerns only what I called 'traditional' positions, i.e., positionsthe identification of which is required in traditional interaction, although it isprobable that such cognitive mechanisms apply to other domains of socialcategorisation.

Conclusions and programme1 This is necessarily a very brief presentation; Chafe's and Goody's hypotheses are

in fact much more refined. The psychological differences studied by Hildyard andOlson (1982) concern both understanding and memorisation. Oral communi-cation allows better recall, and written communication better recognition of asingle text. 'The listeners pay primary attention to the theme of the story, buildinga coherent representation of what was meant. The readers, on the other hand, areable to pay closer attention to the meaning of the sentences per se, recalling moreincidental explicit details' (Hildyard and Olson 1982: 31).

2 For certain limited tasks like palace bookkeeping, pictographic or ideographicsystems seem much more economical than phonetic scripts. When the only use ofliteracy is to count measures of corn and oil in stores, using pictograms is morerational than transcribing the words 'oil' or 'corn' and the words for numbers.This is probably the simplest explanation of the persistence of pictographicelements and ' rebus' interpretations in phonetic scripts, and of the coexistence ofboth types of scripts, like in Mycenean civilisation, with a pictograph and asyllabary used together.

3 This of course is entirely different from the question of 'written traditions',which is outside the scope of this volume. One must notice that the introductionof writing in oral traditions does not create 'written traditions' out of them.

129


4 Obviously, this extension of written rational argumentation does not preclude, asthe Greek case shows eloquently, the development of mystic cults and irrationalsects.

5 The notion of ' claims to truth' (and of competing claims) is inspired by Lloyd1979.

6 Varying degrees of commitment are often described in probabilistic terms, as afull belief in the fact that the considered proposition has a probability 1/x of beingtrue. The application of such ideas, and of refined ' Bayesian' computations, toreal inference processes is not altogether easy, however: see for instance Miller(1987: 297-347) for a discussion from a philosophical viewpoint.

7 The distinction between 'central processes' (such as attention or memorisation)and ' modular processes' (specialised in perceptual information, language parsing,etc.) was introduced by Fodor (1983), together with the idea that, while cognitivepsychology is increasingly successful at describing modular functioning, it hasvery little hope of ever explaining how central processes work.

8 It is not possible here to examine to what extent naturalistic data are an obstacleto precise and testable arguments. Hutchins's study of land-tenure litigation inTrobriand (1980) is a vivid counter-example, which shows that a successfulcognitive formalisation can be based on classical fieldwork data. Hutchins himselfdiscusses the problem in his conclusion (1980: 125-28).

130

Bibliography

Adler, A. and A. Zempleni, 1972. Le Baton de V Aveugle, Paris: Editions Hermann.Agar, M., 1974. Talking about doing. Lexicon and event, Language and Society, 3:

83-89.Ahern, E. M., 1979. The problem of efficacy. Strong and weak illocutionary acts,

Man (n.s.), 14: 1-17.1982. Rules in oracles and games, Man (n.s.), 17: 302-12.

Anderson, J. R. and B. H. Ross, 1980. Evidence against a semantic/episodicdistinction, Journal of Experimental Psychology (Human Learning and Memory),6: 441-65.

Anglin, J. M., 1970. The Growth of Word-meaning, Cambridge: The MIT Press.1977. Word, Object and Conceptual Development, New York: Norton.

Aquili, E. G., C. D. Laughlin and J. McManus, 1979. The Spectrum of Ritual. ABiogenetic Structural Analysis, New York: Columbia University Press.

Atran, S., 1987. Ordinary constraints on the semantics of living kinds. Acommonsense alternative to recent treatments of natural-object terms, Mind andLanguage 2: 27-63.

Austin, J. L., 1962. How To Do Things With Words, Oxford: Clarendon Press.Barnes, J. A., 1967. Politics In A Changing Society, Manchester: Manchester

University Press.Barth, F., 1975. Ritual and Knowledge Among The Bakataman Of New Guinea,

Newhaven/Oslo: Yale University Press/Universitet-forlaget.Bartlett, F. C , 1923. Psychology And Primitive Culture, Cambridge: Cambridge

University Press.1932. Remembering. A Study In Experimental And Social Psychology, Cambridge:

Cambridge University Press.1958. Thinking. An Experimental And Social Study. New York: Basic Books.

Bauer, D. F. and J. Hinnant, 1980. Normal and revolutionary divination. A Kuhnianapproach to African traditional thought, in I. Karp and C. S. Bird (eds.),Explorations In African Systems Of Thought, Bloomington: Indiana UniversityPress.

Beaugrande, R. de, 1982. Story of grammar and grammar of stories, Journal ofPragmatics, 6: 383-422.

Bellman, B. L., 1984. The Language Of Secrecy. Symbols And Metaphors In PoroRitual, New Brunswick: Rutgers University Press.

Bendix, R. and R. Lipset, 1966. Class, Status and Power. Social Stratification inComparative Perspective, New York: The Free Press.

Berlin, B., D. Breedlove and P. Raven, 1973. General principles of classification andnomenclature in folk-biology, American Anthropologist, 75: 214-42.

131

Bibliography

Berreman, G. D., 1968. The concept of 'caste', International Encyclopaedia of theSocial Sciences, New York: McMillan and The Free Press.

Bloch, M., 1974. Symbols, song, dance and features of articulation. Is religion anextreme form of traditional authority?, European Journal of Sociology, 15:55-81.

1985. From cognition to ideology, in R. Fardon (ed.),Power and Knowledge.Anthropological and Sociological Approaches, Edinburgh: Scottish AcademicPress.

Bohannan, L., 1952. A genealogical charter, Africa, 22: 301-15.Bottero, J., 1987. Mesopotamie, Paris: Editions Gallimard.Bowerman, M., 1978. The acquisition of word-meaning: an investigation into some

current conflicts, in N. Waterson and C. Snow (eds.), The Development ofCommunication, New York: Wiley.

1982. Reorganizational processes in lexical and syntactic development, in E.Wanner and L. R. Gleitman (eds.), Language Acquisition: The State of the Art,Cambridge: Cambridge University Press.

Boyer, P., 1980. Les Figures du savoir initiatique, Journal des Africanistes, 50: 31-57.1982a. Recit epique et tradition, UHomme, Revue Francaise dAnthropologie, 27:

5-34.1982b. Le Status des forgerons et ses justifications symboliques, Africa, 53: 44-63.1984. La Tradition comme genre enonciatif, Poetique, 58: 55-72.1986. The 'empty' concepts of traditional thinking. A semantic and pragmatic

description, Man (n.s.), 21: 50-64.1987. The stuff 'traditions' are made of. On the implicit ontology of an

ethnographic category,Philosophy of the Social Sciences, 17: 49-65.1988. Barricades mysterieuses et pieges apense'e, Paris: Societe d'Ethnologie.1989a. Pseudo-kinds and Complex Cultural Categories, unpubl. ms.1989b. Causal Thinking and Its Anthropological Misrepresentation, unpubl. ms.

Boyd, R., 1980. Materialism without reductionism. What physicalism does not entail,in N. Block (ed.), Redings in the Philosophy of Psychology, Cambridge: The MITPress.

Brand, M., 1976. Introduction: defining 'causes', in M. Brand (ed.), The Nature ofCausation, Chicago: University of Illinois Press.

Carey, S., 1978. The child as word learner, in M. Halle and G. A. Miller (eds.),Linguistic Theory And Psychological Reality, Cambridge: The MIT Press.

1982. Semantic development: the state of the art, in E. Wanner and L. R. Gleitman(eds.), Language Acquisition: The State of the Art, Cambridge: CambridgeUniversity Press.

1985. Conceptual Change in Childhood, Cambridge: The MIT Press.Carnap, R., 1956. The methodological character of theoretical concepts, in H. Feigl

and M. Scriven (eds.), Minnesota Studies in the Philosophy of Science (I),Minneapolis: University of Minnesota Press.

Chafe, W., 1982. Integration and involvement in speaking, writing and oral literature,in D. Tannen (ed.), Spoken and Written Language, Norwood: Ablex.

Churchland, P. M., 1985. Conceptual progress and word/world relations. In searchof natural kinds, Canadian Journal of Philosophy, 15: 1-17.

Cicourel, A., 1972. Basic and normative rules in the negociation of status and role,in D. Sudnow (ed.), Studies in Social Interaction, New York: The Free Press.

Clark, E. V., 1973. Non-linguistic strategies and the acquisition of word-meaning,Cognition, 2: 161-82.

Clement, P., 1948. Le Forgeron en Afrique Noire. Quelques attitudes du groupe a sonegard, Revue de Geographie Humanie et dEthnographie, 2: 1-15.

132

Bibliography

Cole, M. and S. Scribner, 1974. Culture And Thought. A Psychological Introduction,New York: John Wiley and Sons.

Cole, M., et al., 1971. The Cultural Context of Learning and Thinking. An Explorationin Experimental Anthropology, London: Methuen.

Craik, F. I. M. (ed.), 1979. Levels Of Processing In Human Memory, Hillsdale:Lawrence Erlbaum Associates.

Davidson, D., 1980. Essays On Actions And Events, Oxford: Clarendon Press.Detienne, M., 1966. Les Maitres de ve'rite dans la Grece archaique, Paris: Francois

Maspero.Dieterlen, G. and Y. Cisse, 1972. Les Fondements de la Societe d Initiation du Komo,

Paris/The Hague: Mouton and Co.Dougherty, J. D. W. (ed.), 1985. Directions in Cognitive Anthropology, Urbana/

Chicago: University of Illinois Press.Douglas, M., 1982. In the Active Voice, London: Routledge and Kegan Paul.Dretske, F., 1981. Knowledge and the Flow of Information, Oxford: Basil Blackwell.Eisenstadt, S. N., 1956. From Generation to Generation, London: The Free Press.Evans, G., 1982. The Varieties of Reference, Oxford: Clarendon Press.Evans-Pritchard, E. E., 1937. Witchcraft, Oracles and Magic among the Azande,

Oxford: Clarendon Press.Favret Saada, J., 1980. Deadly Words, Cambridge/Paris: Cambridge University

Press/Maison des Sciences de FHomme.Favret Saada, J. and J. Contreras, 1984. Comment produire de Fenergie avec deux

jeux de cartes, Bulletin d Ethnomedecine 24.Fernandez, J. W., 1982. Bwiti. An Ethnography of the Religious Imagination in Africa,

Princeton: Princeton University Press.1986. Persuasions and Performances. The Play of Tropes in Culture, Bloomington:

Indiana Unversity Press.Finnegan, R., 1969. How to do things with words: performative utterances among

the Limba of Sierra Leone, Man (n.s.), 4: 57-51.Fodor, J. A., 1981. Representations. Philosophical Essays on the Foundations of

Cognitive Science, Cambridge: The MIT Press.1983. The Modularity Of Mind. An Essay On Faculty Psychology, Cambridge: The

MIT Press.Fodor, J. A. et al, 1980. Against definitions, Cognition, 8: 263-367.Fodor, J. D. et al., 1975. The psychological unreality of semantic representations,

Linguistic Inquiry, 6: 515—31.Fogelson, R. D. and R. N. Adams (eds.), 1977. The Anthropology of Power, New

York: Academic Press.Fortes, M., 1966. Religious premises and logical technique in divinatory ritual,

Philosophical Transactions of the Royal Society of London, 251: 409-22.Gardner, D. S., 1983. Performativity and ritual: the Mianmin case, Man (n.s.), 18:

346-60.Gasking, D., 1955. Causation and recipes. Mind, 64: 479-87.Gellner, E., 1974. The Legitimation of Belief Cambridge: Cambridge University

Press.1979. Spectacles and Predicaments. Essay in Social Theory, Cambridge: Cambridge

University Press.Gelman, S., 1988. The development of induction within natural kind and artefact

categories, Cognitive Psychology, 20: 65-95.Gelman, S. and E. Markman, 1987. Young children's inductions from natural kinds:

the role of categories and appearances, Child Development, 58: 32-41.

133

Bibliography

Genest, S., 1976. La Transmission du savoir chez les Forgerons Mafa (Nord-Cameroun), Ph.D. Dissertation, University of Paris-Sorbonne.

Goodenough, W., 1965. Rethinking 'status' and 'role', in M. Banton (ed.), TheRelevance of Models for Social Anthropology, London: Tavistock Publications.

Goodman, N., 1954. Fact, Fiction and Forecast, Cambridge: Harvard UnversityPress.

Goody, J. R., 1977. The Domestication of the Savage Mind, Cambridge: CambridgeUniversity Press.

1986. The Logic of Writing and the Organisation of Society, Cambridge: CambridgeUniversity Press.

1987. The Interface of the Oral and the Written, Cambridge: Cambridge UniversityPress.

Goody, J. R. and I. Watt, 1968. The consequences of literacy, in J. R. Goody (ed.),Literacy in Traditional Societies, Cambridge: Cambridge University Press.

Gutmann, B., 1926. Das Recht der Dschagga, Munich.Halbwachs, M., 1925. Les Cadres sociaux de la memoire, Paris: Felix Alcan.Halliday, M. A. K., 1976. Anti-language, American Anthropologist, 78: 570-84.Hallpike, C. R., 1979. The Foundations of Primitive Thought, Oxford: Clarendon

Press.Hart, H. L. A. and T. Honore, 1959. Causation in the Law, Oxford: Clarendon Press.Herdt, G. (ed.) 1982. Rituals of Manhood. Male Initiation in Papua New Guinea,

Berkeley: University of California Press.Hildyard, A. and T. R. Olson, 1982. On the comprehension and memory of oral vs.

written discourse, in D. Tannen (ed.), Spoken and Written Language. ExploringOrality and Literacy,Norwood: Ablex.

Hirschfeld, L. A., 1988. On acquiring social knowledge, Man, 23.Hollis, M., 1970. Reason and ritual, in B. Wilson (ed.), Rationality, Oxford: Basil

Blackwell.Hollis, M., and S.Lukes (eds.), 1982. Rationality and Relativism, Oxford: Basil

Blackwell.Holy, L. and M. Stuchlik, 1983. Actions, Norms and representations. Foundations of

Anthropological Inquiry, Cambridge: Cambridge University Press.Horton, R., 1967a, b. African traditional thought and modern science, Africa, 38:

50-71 and 155-87.1970. African traditional thought and modern science, in B. Wilson (ed.),

Rationality, Oxford: Basil Blackwell.1982. Tradition and modernity revisited, in M. Hollis and S. Lukes

(eds.), Rationality and Relativism, Oxford: Basil Blackwell.Houseman, M., 1984. Les Artifices de la logique initiatique, Journal des Africanistes,

54: 41-65.1986. Le Mai pour le Male, un bien initiatique, in J. Hainard and R. Kaehr (eds.),

Le Mai et la douleur, Neuchatel: Institut d'Ethnologie.Humphrey, C , 1976. Omens and their explanation among the Buryat, European

Journal of Sociology, 17: 21-38.Hutchins, E., 1980. Culture and Inference. A Trobriand Case Study, Cambridge:

Harvard University Press.1981. Reasoning in Trobriand discourse, in R. W. Casson (ed.), Language, Culture

and Cognition, New York: Macmillan.Hymes, D., 1974. Ways of speaking, in R. Bauman and J. Sherzer (eds.), Explorations

in the Ethnography of Speaking, Cambridge: Cambridge University Press.Ingold, T., 1986. Evolution and Social Life, Cambridge: Cambridge University Press.

134

Bibliography

Jakobson, R., 1966. Retrospect, in R. Jakobson, Selected Writings IV: Early SlavicPaths and Crossroads, The Hague/New York: Mouton.

1981. Grammatical parallelism and its Russian facet, in R. Jakobson, SelectedWritings HI: Poetry of Grammar and Grammar of Poetry, The Hague/NewYork: Mouton.

Katz, B., G. Baker and J. MacNamara, 1974. What's in a name? On the child'sacquisition of proper and common nouns, Child Development, 45: 269-73.

Katz, J. J., 1977. Propositional Structure andIllocutionary Force, New York: ThomasY. Crowell.

Keesing, R., 1984. Rethinking Mana, Journal of Anthropological Research, 40:137-56.

1985. Conventional metaphors and anthropological metaphysics: the problematicof cultural translations, Journal of Anthropological Research, 41: 201-17.

Keil, F. C , 1979. Semantic and Conceptual Devlopment, Cambridge: HarvardUniversity Press.

1986. The acquisition of natural kind and artefact terms, in A. Marrar and W.Demopoulos (eds.), Conceptual Change, Norwood: Ablex.

Kiparsky, P., 1976. Oral poetry: some linguistic and typological considerations, in B.A. Stolz and R. S. Shannon (eds.), Oral Literature and the Formula, Ann Arbor:Michigan University Press.

La Fontaine, J. S., 1977. The power of rights, Man (n.s.), 12: 421-37.1985. Initiation. Ritual Drama and Secret Knowledge Across the World, Harmonds-

worth: Penguin Books.Lakoff, G. and M. Johnson, 1977. Metaphors We Live By, Chicago: University of

Chicago Press.Lakoff, G. and Z. Kovecses, 1986. The cognitive model of anger inherent in English,

in D. Holland and N. Quinn (eds.), Cultural Models in Language and Thought,New York: Cambridge University Press.

Laurendau, M. and A. Pinard, 1962. Causal Thinking in the Child, New York:International Universities Press.

Lee, D. D., 1949. Being and value in a primitive culture, Journal of Philosophy, 8(13):401-15.

Leslie, A., 1979. The Representation of Perceived Causal Connection, D.Phil, thesis,University of Oxford.

Levinson, S., 1983. Pragmatics, Cambridge: Cambridge University Press.Levi-Strauss, C , 1987. Introduction to the Work of Marcel Mauss, London:

Routledge and Kegan Paul.Lewis, G. A., 1980. Day of Shining Red. An Essay on Understanding Ritual,

Cambridge: Cambridge University Press.Lloyd, G. E. R., 1979. Magic, Reason and Experience, Cambridge: Cambridge

University Press.Loewe, M. and C. Blacker (eds.), 1981. Divination and Oracles, London: George

Allen and Unwin.Luhrmann, T., 1986. Witchcraft, morality and magic in contemporary London,

International Journal of Moral and Social Sciences, 1: 77-94.Luria, A., 1976. Cognitive Development. Its Cultural and Social Foundations,

Cambridge: Harvard University Press.Luther, W., 1935. 'Wahrheit' und 'Liige' im altesten Griechentum, Gottingen:

Georg August Universitat (Inaugural-Dissertation).MacCulloch, J. A., 1912. Heart, in J. Hastings (ed.), Encyclopaedia of Religion and

Ethics, Edinburgh: T. and T. Clark.Mackie, J. L., 1974. The Cement of the Universe, Oxford: Clarendon Press.

135

Bibliography

MacNamara, J., 1982. Names for Things. A Study of Human Learning, Cambridge:The MIT Press.

Mallart-Guimera, L., 1975. Ni dos ni ventre: magie, sorcellerie et religion evuzok,UHomme. Revue Francaise dAnthropologie, 15: 35-65.

1981. Ni dos ni ventre, Paris: Societe d'Ethnographie.Maret, P. de, 1980. Ceux qui jouent avec le feu: la place du forgeron en Afrique

centrale, Africa, 50: 263-79.Martin, J. Y., 1970. Les Matakam du Nord-Cameroun. Dynamismes sociaux et

problemes de modernisation, Paris: Orstom.Mauss, M., 1951. Sociologie et Anthropologie, Paris: Presses Universitaires de France.Mendonsa, E. L., 1976. The Politics of Divination, A Processual View of Illness and

Deviation among the Sisala of Northern Ghana, Berkeley: University of CaliforniaPress, 1982.

1982. Characteristics of Sisala diviners, in A. Bharati (ed.), The Realm of the Extra-Human, Agents and Audiences, Paris/The Hague: Mouton and Cy.

Miller, R., 1987. Fact and Method. Explanation, Confirmation and Reality in theNatural and the Social Sciences, Princeton: Princeton University Press.

Needham, R., 1972. Belief, Language and Experience, Oxford: Basil Blackwell.1976. Skulls and causality, Man (n.s.), 11: 71-88.

Nuchelmans, G., 1973. Theories of the Proposition. Ancient and Medieval Conceptionsof the Bearers of Truth and Falsity, Amsterdam: North-Holland.

1980. Late-Scholastic and Humanist Theories of the Proposition, Amsterdam:North-Holland.

Osherson, D. N., 1978. Three conditions on conceptual naturalness, Cognition, 6:263-89.

Park, G., 1963. Divination and its social contexts, Journal of the Royal An-thropological Institute, 93: 195-209.

Philsooph, H., 1972. Primitive magic and Mana, Man (n.s.), 6: 182-203.Posner, M. I., and S. W. Keele, 1968. On the genesis of abstract ideas, Journal of

Experimental Psychology, 11: 353-63.Putman, H., 1975. Mind, Language and Reality. Philosophical Papers (II),

Cambridge: Cambridge University Press.1983. Realism and Reason. Philosophical Papers (III), Cambridge: Cambridge

University Press.Rappaport, R. A., 1974. The obvious aspects of ritual, Cambridge Anthropology, 2:

3-69.Reynolds, B., 1963. Magic, Divination and Witchcraft among the Barotse of Northern

Ghana, London: Chatto and Windus.Richards, A. I., 1956. Chisunqu, A GirFs Initiation Ceremony in Northern Rhodesia,

London: Faber.Sales, A. de, 1986. Actes et paroles dans les rituels Chamaniques des Kham Magar,

Ph.D. Dissertation, University of Paris-Nanterre.Schustack, M. W. and R. J. Sternberg, 1981. Evaluation of evidence in causal

inference, Journal of Experimental Psychology: General, 110: 101-20.Schwarz, S. P. (ed.), 1977. Naming, Necessity and Natural Kinds, Ithaca: Cornell

University Press.1979. Natural kind terms, Cognition, 7: 301-15.

Scribner, S., 1975. Recall of classical syllogisms. A cross-cultural investigation oferror on logical problems, in R. J. Falmagne (ed.) Reasoning: Representation andProcess, Hillsdale: Lawrence Erlbaum Associates.

Scribner, S. and M. Cole, 1981. The Psychology of Literacy, Cambridge: HarvardUniversity Press.

Severi, C , 1985. Penser par sequences, penser par territoires, Communications, 41:169-90.

136

Bibliography

Forthcoming. The invisible path. Ritual representation of suffering in Cunatraditional thought, Res. Anthropology and Aesthetics, 14: 66-81.

Sherzer, J., 1985. Kuna Ways of Speaking. An Ethnographic Perspective, Austin:University of Texas Press.

Shils, E., 1981. Tradition, London: Faber and Faber.Shoben, E. J., 1984. Semantic and episodic memory, in R. S. Wyer and T. K. Srull

(eds.), Handbook of Social Cognition (II), Hillsdale: Lawrence ErlbaumAssociates.

Skorupski, J., 1976. Symbol and Theory. A Philosophical Study of Theories of Religionin Social Anthropology, Cambridge: Cambridge University Press.

Smith, E. E., and D. L. Medin, 1981. Categories and Concepts, Cambridge: HarvardUniversity Press.

Sommers, F., 1959. The ordinary language tree, Mind, 68: 160-85.Southwold, M., 1979. Religious Belief, Man (n.s.), 14: 628-44.Sperber, D., 1982. Apparently irrational beliefs, in M. Hollis and S. Lukes (eds.),

Rationality and Relativism, Oxford: Basil Blackwell.Steiner, F. B., 1954. Chagga truth. A note on Gutmann's account of the Chagga

concept of truth in 'Das Recht der Dschagga', Africa, 24: 364-9.Stich, S., 1983. From Folk Psychology to Cognitive Science. The Case Against Belief,

Cambridge: The MIT Press.Strathern, M., 1985. Knowing power and being equivocal. Three Melanesian

contexts, in R. Fardon (ed.), Power and Knowledge, Anthropological andSociological Approaches, Edinburgh: Scottish Academic Press.

Strawson, P. F., 1949. Truth, Analysis, 9: 83-97.Tambiah, S. J., 1968. The magical power of words, Man (new series), 3: 175-208.Tannen, D., 1985. Relative focus on involvement in oral and written discourse, in D.

R. Olson, N. Torrance and A. Hildyard (eds.), Literacy, Language and Learning.The Nature and Consequences of Reading and Writing, Cambridge: CambridgeUniversity Press.

Taylor, R., 1963, Causation, The Monist, 47: 287-313.Tessmann, G., 1913. Die Pangwe. Volkerkundliche Monographic eines west-

afrikanischen Negerstammes, Berlin: Ernst Wasmuth.Thomas, K., 1971. Religion and the Decline of Magic. Studies in Popular Beliefs in 16 th

and 17th-century England, London: Weidenfeld and Nicolson.Todd, D. M., 1977. Caste in Africa? Africa, 47 (4): 398^12.Tuden, A. and L. Plotnicov (eds.), 1970. Social Stratification in Africa, New York:

The Free Press.Tulving, E., 1983. Elements of Episodic Memory, Oxford: Clarendon Press.Ullmann, S., 1962. Semantics. An Introduction to the Science of Meaning, New York:

Barnes and Noble.Urban, W. M., 1939. Language and Reality, London: George Allen and Unwin.Vaughan, J. H., 1970. Caste systems in the Western Sudan, in A. Tuden and L.

Plotnicov (eds.), Social Stratification in Africa, New York: The Free Press.Vendler, Z., 1967. Causal relations, Journal of Philosophy, 64: 704-13.Vernant, J. P. (ed.), 1974. Divination et rationalite, Paris: Editions du Seuil.Weil, E., 1971. Tradition et traditionnalisme, in E. Weil, Essais et Conferences (I),

Paris: Librairie Plon.Wilson, B. (ed.), 1980. Rationality, Oxford: Basil Blackwell.Zeitlyn, D., 1987. Mambila divination, Cambridge Anthropology, 12: 20-51.Zempleni, A., 1986. La 'Maladie' et ses 'causes', LEthnographie (new series), 4:

1-34.Zwe Nguema, 1972. Un mvet de Zwe Ngue'ma, Paris: Armand Colin.

137

Index

abele, 48ff.Adams, 123, 133Adler, 70, 131Ahern, 127, 131aletheia, 49, 124Anderson, 124, 131Austin, 86, 131authority, traditional, see truth

Baktaman, 98Barth, 75, 98, 128, 131Bartlett, 122, 131Bauer, 65, 131Beaugrande, 122, 131bekong, 40-41belief: everyday and traditional, 52ff;

theories of, 5Iff., 118, 125Bendix, 128, 131Berlin, 104, 131Berreman, 129, 131Beti, 96-97black box, see psychologyBlacker, 65, 135Bloch, 82-6, 89-90, 127-28, 132Bottero, 65, 132Buryat, 126

Carey, 124, 126, 132Caste, 101-2, 129Categories: acquisition of, 34ff., 43-44, 124;

anchoring in episodes, 41^44, 109; shareddefinitions of, 25, 37ff; 'empty', 24ff.;and natural kinds, 38ff., 123; non-definability, 24, 37, 123; as 'signifiers',27ff., 41; as theoretical terms, 29ff., 41

causality, 75-78, 126Chafe, 115, 129, 132Churchland, 123, 132Cicero, 62, 68, 125Cisse, 98, 132Clement, 101, 132cohesiveness, 3

Cole, 116, 132collective representations, 4communication, 20, 109competence, 94, 99-101, 111-12concepts, see categoriesconservation, 8conservatism, 4Contreras, 67, 133conventional metaphor, 50-51, 123, 124Cuna, 80-81, 89customisation: in general, 60; of persons,

94-103; of speech in divination, 61-78; ofspeech in ritual 79-93

Detienne, 124, 133Dieterlen, 98, 133diil, 97-98discourse registers: about evitr, 30ff.; expert

and common, 33-34; and guarantees oftruth, 33, 44-45

divination, 46, 61-78: causal hypothesis on,72ff.; intentional view of, 65ff., 125; asmanipulation, 61ff., 125; semiotic view of,68ff., 126; types of, 63ff., 126

Douglas, 122, 133Dretske, 127, 133duwaiin, 97-98

energy, 'mystical', 26episodic memory, 42ff.Evans, 126, 133Evans-Pritchard, vii, 62, 133event-talk, 12, 42evur, 25ff.

Fang: literary traditions, 6-11, 20-21, 121;witchcraft and magic, 25ff., 34ft, 123;notions of truth, 48, 52

Favret-Saada, 67, 133Fernandez, 84, 124, 133Finnegan, 127, 133Fodor, J. A., 123, 130, 133

138

Index

Fodor, J. D., 123, 133Fogelson, 123, 133force, 'vital', 26formalised speech, 82ff.Fortes, 66, 71, 133

Gardner, 97-8, 127, 133Gellner, 113, 122, 133Gelman, 104, 133Genest, 102, 133Gisu, 96Goodenough, 128, 134Goody, 114-15, 129, 134great divide, see tradition, definitions ofGutmann, 49, 124, 134

Halbwachs, 122, 134Halliday, 127, 134Hart, 126, 134Herdt, 128, 134Hildyard, 115, 129, 134Hinnant, 65, 132Hocart, 129Hollis, 52, 56, 134holy, 121-22, 134Honore, 126, 134Horton, 4-5, 28, 121, 134Houseman, 96, 128, 134Humphrey, 126, 134Hutchins, 130, 134

iconicity, 88, 128illocutionary force, 82-83, 85-86, 128iloha, 49Ingold, 18, 122, 134initiation: and causal criteria, 97ff.; and

knowledge, 95ff., 128interaction, representation of, 18-20

Jakobson, 128, 134

Katz, 128, 135Keele, 124, 136Keesing, 27, 123-24, 135Keil, 125, 129, 135Kham-Magar, 87Kiparsky, 128, 135

La Fontaine, 96, 128, 135Lakoff, 124, 135Laurendeau, 126, 135Leslie, 126, 135Levi-Strauss, 27, 135Levinson, 128, 135Lewis, 3, 135

Lipset, 128, 131literacy, 114-16, 129literalism, 11-12, 17, 92-93living kinds, see natural kind termsLloyd, 116, 130, 135Loewe, 65, 135lohU 49Luhrmann, 112, 135Lukes, 52, 135Luther, 124, 135

machin, 27-28Mackie, 126, 135MacNamara, 124, 135Madagascar, 82ff., 89-90magic, 127Malinowski, viiMallart-Guimera, 35, 123-4, 136Mambila, 71mana, 27, 124Maret, 101, 136Markman, 104, 133Martin, 102, 136Mauss, 27, 136Medin, 123, 137meduk, 48ff.memory, 16-17, 41-44, 122, 124Mendonsa, 62, 70, 136Merina, 82ff., 89-90Mianmin, 97-98Miller, 130, 136Mundang, 70ff.mvet, 6-8

natural kind terms, 38, 104-5, 123, 129Navajo, viiNeedham, 76, 125, 136neo-intellectualism, 4-5, 121ngengang, 34ff.Nuer, vii, 53

Olson, 115, 129, 134ostension, 36ff.

Paracelsus, 126perlocution, 86, 128Pinard, 126, 135Plotnicov, 129, 137Plutarch, 125-26positions: naturalised, 103-6, 109; and

occupation, 101-5, 128, 129; and truth,94ff., 99-101

Posner, 124, 136propositional potential, 82ff., 127psychology: ad-hoc versions of, ix; and

139

anthropology, ix; black box arguments, ixPutnam, 123, 136

Rappaport, 79, 136rationality debate, 5Iff.repetition, 1-3, 121; automatic, 13, 16,

18-20; intentional, 13-14; and literalism,92-93

representations, shared, 20ff., 32ff., 42Reynolds, 65, 136Richards, 95, 136Ritual speech: characteristics, 79-93, 127;

and truth, 88ff., 127Ross, 124, 131

sales, 87, 136salience: of claims to truth, 116-19, 130; of

events in traditions, 2Schustack, 126, 136Schwarz, 123, 136Scribner, 116, 132, 136semantic memory, 42Sentences, vs. utterances, 86, 91-93Severi, 127, 136Shaman, 1Sherzer, 80ff., 89, 137Shils, vii, viii, 137Sisala, 70ff.Skorupski, 4, 137Smith, 123, 137Southwold, 53, 137speech-acts, 127Sperber, 53, 137stereotype, 38Sternberg, 126, 136Strawson, 125, 137structuralism, 27ff.Stuchlik, 121-22, 134surface properties, 13-14

Tallensi, 71Tambiah, 127, 137Tannen, 115, 137Tessmann, 122, 137theories, see tradition, definitions ofThomas, 62, 125, 137Todd, 129, 137Tradition, definitions of: common

anthropological, 3ff., 10ff., 22, 107-8, 117;commonsense and scientific, viii; and'modernity', 107, 110-14; as the past,viii; as cultural 'theories', 4-5, 12-13;used in this book, viii, 1-3, 23

traditionalism, see conservatismTrobriand, vii

Index

true, 27-28Truth: and authority, 58; causal criteria of,

72ff., 88ff., 109, 126; conceptions of, 55ff.;criteria of, 55ff.; guarantees of, 33, 92-93;and positions, 94ff., 99-101, 105-6;problems about, 46ff.; and situations,59-60; terms and metaphors, 50-51, 124;terms and predicates, 48ff., 124-25

Tuden, 129, 137Tulving, 124, 137

utterances as events, 86ff., 91-93

Vaughan, 129, 137

Watt, 114, 134Wilson, 52, 137witchcraft: Fang, 25fT., 31, 34^36; London,

112world-views, see tradition, definitions of

Zande, vii, 62Zeitlyn, 65, 71, 137Zempleni, 66, 70, 131, 137Zwe Nguema, 7, 122, 137

140

Cambridge Studies inSocial AnthropologyEditor: JACK GOODY

1 The Political Organisation of UnyamweziR. G. ABRAHAMS

2 Buddhism and the Spirit Cults in North-East Thailand*S. J. TAMBIAH

3 Kalahari Village Politics: An African DemocracyADAM KUPER

4 The Rope of Moka: Big-Men nd Ceremonial Exchange in Mount Hagen, New GuineaANDREW STRATHERN

5 The Majangir: Ecology and Society of a Southwest Ethiopian PeopleJACK STAUDER

6 Buddhist Monk, Buddhist Layman: A Study of Urban Monastic Organisation inCentral Thailand

JANE BUNNAG7 Contexts of Kinship: An Essay in the Family Sociology of the Gonja of Northern

GhanaESTHER N. GOODY

8 Marriage among a Matrilineal Elite: A Family Study of Ghanaian Civil ServantsCHRISTINE OPPONG

9 Elite Politics in Rural India: Political Stratification and Political Alliances in WesternMaharashtra

ANTHONY T. CARTER10 Women and Property in Morocco: Their Changing Relation to the Process of Social

Stratification in the Middle AtlasVANESSA MAHER

11 Rethinking SymbolismDAN SPERBER, translated by Alice L. Morton

12 Resources and Population: A Study of the Gurungs of NepalALAN MACFARLANE

13 Mediterranean Family StructuresEDITED BY J. G. PERISTIANY

14 Spirits of Protest: Spirit-Mediums and the Articulation of Consensus among the Zezuruof Southern Rhodesia (Zimbabwe)

PETER FRY15 World Conqueror and World Renouncer: A Study of Buddhism and Polity in Thailand

against a Historical Background*S. J. TAMBIAH

16 Outline of a Theory of Practice*PIERRE BOURDIEU, translated by Richard Nice

17 Production and Reproduction: A Comparative Study of the Domestic DomainJACK GOODY

18 Perspectives in Marxist AnthropologyMARUCIE GODELIER, translated by Robert Brain

141

19 The Fate Shechem, or the Politics of Sex: Essays in the Anthropology of theMediterranean

JULIAN PITT-RIVERS20 People of the Zongo: The Transformation of Ethnic Identities in Ghana

ENID SCHILDKROUT21 Casting out Anger: Religion among the Taita of Kenya

GRACE HARRIS22 Rituals of the Kandyan State

H. L. SENEVIRATNE23 Australian Kin Classification

HAROLD W. SCHEFFLER24 The Palm and the Pleiades: Initiation and Cosmology in Northwest Amazonia

STEPHEN HUGH-JONES25 Nomads of Southern Siberia: The Pastoral Economies of Tuva

s. i. VANSHTEIN, translated by Michael Colenso26 From the Milk River: Spatial and Temporal Processes in Northwest Amazonia

CHRISTINE HUGH-JONES27 Day of Shining Red: An Essay on Understanding Ritual

GILBERT LEWIS28 Hunters, Pastoralists and Ranchers: Reindeer Economies and their Transformations

TIM INGOLD29 The Wood-Carvers of Hong Kong: Craft Production in the World Capitalist Periphery

EUGENE COOPER30 Minangkabau Social Formations: Indonesian Peasants and the World Economy

JOEL S. KAHN31 Patrons and Partisans: A Study of Two Southern Italian Communes

CAROLINE WHITE32 Muslim Society*

ERNEST GELLNER33 Why Marry Her? Society and Symbolic Structures

LUC DE HEUSCH, translated by Janet Lloyd34 Chinese Ritual and Politics

EMILY MARTIN AHERN35 Parenthood Social Reproduction: Fostering and Occupational Roles in West Africa

ESTHER N. GOODY36 Dravidian Kinship

THOMAS R. TRAUTMANN37 The Anthropological Circle: Symbol Function, History*

MARC AUGE, translated by Martin Thorn38 Rural Society in Southeast Asia

KATHLEEN GOUGH39 The Fish-People: Linguistic Exogamy and Tukanoan Identity in Northwest Amazonia

JEAN E. JACKSON40 Karl Marx Collective: Economy, Society and Religion in a Siberian Collective Farm*

CAROLINE HUMPHREY41 Ecology and Exchange in the Andes

edited by DAVID LEHMANN42 Traders without Trade: Responses to Trade in two Dyula Communities

ROBERT LAUNAY43 The Political Economy of West African Agriculture*

KEITH HART44 Nomads and the Outside World

A. M. KHAZANOV, translated by Julia Crookenden45 Actions, Norms and Representations: Foundations of Anthropological Inquiry*

LADISLAV HOLY AND MILAN STUCHLIK46 Structural Models in Anthropology*

PER HAGE AND FRANK HARARY

142

47 Servants of the Goddess: The Priests of a South Indian Templec. J. FULLER

48 Oedipus and Job in West African Religion*MEYER FORTES

49 The Buddhist Saints of the Forest and the Cult of Amulets: A Study in Charisma,Hagiography, Sectarianism, and Millennial Buddhism*

S. J. TAMBIAH50 Kinship and Marriage: An Anthropological Perspective (available in paperback/in the

USA only)ROBIN FOX

51 Individual and Society in Guiana: A Comparative Study of Amerindian SocialOrganization*

PETER RIVIERE52 People and the State: An Anthropology of Planned Development*

A. F. ROBERTSON53 Inequality among Brothers; Class and Kinship in South China

RUBIE S. WATSON54 On Anthropological Knowledge*

DAN SPERBER55 Tales of the Yanomami: Daily Life in the Venezuelan Forest*

JACQUES LIZOT, translated by Ernest Simon56 The Making of Great Men: Male Domination and Power among the New Guinea

Baruya*MAURICE GODELIER, translated by Rupert Swyer

57 Age Class Systems: Social Institutions and Polities Based on Age*BERNARDO BERNARDI, translated by David I. Kertzer

58 Strategies and Norms in a Changing Matrilineal Society: Descent, Succession andInheritance among the Toka of Zambia

LADISLAV HOLY59 Native Lords of Quito in the Age of the Incas: The Political Economy of North-

Andean ChiefdomsFRANK SALOMON

60 Culture and Class in Anthropology and History: A Newfoundland IllustrationGERALD SIDER

61 From Blessing to Violence: History and Ideology in the Circumcision Ritual of theMerina of Madagascar*

MAURICE BLOCH62 The Huli Response to Illness

STEPHEN FRANKEL63 Social Inequality in a Northern Portuguese Hamlet: Land, Late Marriage, and

Bastardy, 1870-1978BRIAN JUAN O'NEILL

64 Cosmologies in the Making: A Generative Approach to Cultural Variation in InnerNew Guinea

FREDRIK BARTH

65 Kinship and Class in the West Indies: A Genealogical Study of Jamaica and GuyanaRAYMOND T. SMITH

66 The Making of the basque NationMARIANNE HEIBERG

67 Out of Time: History and Evolution in Anthropological Discourse

NICHOLAS THOMAS

*AVAILABLE IN PAPERBACK

143

[Pascal Boyer] Tradition as Truth and Communicatio

Documents

Transcript of [Pascal Boyer] Tradition as Truth and Communicatio